Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
WenLi
Efficiency of
Manufacturing
Processes
Energy and Ecological Perspectives
Sustainable Production, Life Cycle Engineering
and Management
Series editors
Christoph Herrmann, Braunschweig, Germany
Sami Kara, Sydney, Australia
Modern production enables a high standard of living worldwide through products
and services. Global responsibility requires a comprehensive integration of sustain-
able development fostered by new paradigms, innovative technologies, methods and
tools as well as business models. Minimizing material and energy usage, adapting
material and energy flows to better fit natural process capacities, and changing con-
sumption behaviour are important aspects of future production. A life cycle perspec-
tive and an integrated economic, ecological and social evaluation are essential re-
quirements in management and engineering. This series will focus on the issues and
latest developments towards sustainability in production based on life cycle thinking.
Efficiency of Manufacturing
Processes
Energy and Ecological Perspectives
13
Wen Li
Sustainable Manufacturing and Life Cycle
Engineering
University of New South Wales
Sydney
Australia
vii
viii Foreword
In addition, Dr. Li has further decomposed the derived SEC models of manu-
facturing processes in order to explore the mechanism of the model coefficients. A
clustered model has been also generated for a rough estimation. In order to over-
come the high efforts of empirical modelling, the methodology has been m odified
for the implementation in an industrial environment. Moreover, the derived
SEC models can be used for improving the quality of life cycle inventory (LCI)
databases. As a result, it encourages progress in industry and research towards
sustainability in manufacturing.
The present book has been developed mainly in the context of my work as a
researcher within the Sustainable Manufacturing and Life Cycle Engineering
Research Group at The University of New South Wales (SMLCE@UNSW). First
and foremost, I owe my sincerest gratitude to my supervisor and mentor, Prof.
Sami Kara, who has supported me throughout my research with his patience and
knowledge. It was always joyful and fruitful to discuss research questions as well
as philosophical topics with him during all kinds of occasions. I greatly appreciate
the excellent example he has provided as a successful academic. One simply could
not wish for a better or friendlier supervisor.
I would also offer my sincere gratitude to Prof. Dr.-Ing. Christoph Herrmann
in Sustainable Manufacturing and Life Cycle Engineering Research Group at
Technische Universitt Braunschweig (TUBS). During my research visit to TUBS,
I have been well supported and warmly welcomed by Prof. Herrmann and his
research team. He has not only provided constructive suggestions for my research,
but also integrated me as a member in his research team.
I am indebted to many of my colleagues in the SMLCE@UNSW research
group. Their encouragement and companionship has been very important to me.
I would like to particularly acknowledge Dr. Seung Jin Kim and Dr. Supachai
Vongbunyong for the friendship that I am keen to maintain. It was also a great
pleasure to work with colleagues from the TUBS team, particularly Dr.-Ing.
Sebastian Thiede, Dr.-Ing. Andr Zein, Dr.-Ing. Gerrit Posselt and Mr. Marius
Winter, Dr.-to-be. Their enthusiasm, professionalism and knowledge for the
research of sustainable manufacturing have deeply encouraged me for my
research. Special thanks go to Ms Anne-Marie Schlake who heartened me to finish
this book.
I greatly appreciate Advanced Manufacturing CRC (AMCRC) for funding
this research. Also, I have received great supports from the School of Mechanical
and Manufacturing Engineering, UNSW, especially from Mr. Martyn Sherriff for
assisting me with all the experiments. I would like to extend my gratitude to all the
technicians from Sydney City TAFE who has also provided voluntary assistance
for this research.
ix
x Acknowledgments
Last but not least, I would like to attribute this book to my family for their
love and encouragement. Most of all is for my loving, supportive and encourag-
ing wife, Jie Zhu, whose faithful support is so appreciated. Equally important
and appreciated is my daughter, Lynn Li, who is the true driving force for me to
accomplish this book. Also, I thank my parents Rong Han and Fangze Li who
have unconditionally supported me in all my pursuits. I would like to further thank
Jies parents who have provided me with unending encouragement and support.
1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Problem Statement and Scope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research Objective and Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . 2
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
xi
xii Contents
7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2 Research Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.3 Recommendations for Future Works. . . . . . . . . . . . . . . . . . . . . . . . . 145
Appendix 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Appendix 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Appendix 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Appendix 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Chapter 1
Introduction
The global population has already exceeded 7 billion in 2011, and the growth rate
is not evidently slowing down in the coming years (UNFPA 2011). Pressure from
the population and increasing demand has urged an important question that needs
to be answered by politicians, economists, scientists, and engineers: how can our
planet be sustainable? Although reducing population or curbing demand is infea-
sible in any near future, all types of resources can be used more efficiently, i.e.
energy and eco-efficiency.
Manufacturing is one of the major energy and resource end-users compared to
other sectors such as transport, households, services and others. Globally, this sec-
tor consumed the largest share of total energy generated in 2005 (33 %), with the
corresponding highest proportion of CO2 emissions (38 %), (IEA 2008). In addi-
tion, the energy price has increased sharply as well as the resource price for pro-
ducing energy such as fossil fuels, coal, natural gas, and crude oil. Consequently,
energy costs have become a topic high on the agenda for manufacturing enter-
prises. Therefore, improving energy and eco-efficiency can reduce not only these
costs but also the associated environmental impacts. This is particularly important
for those countries having high carbon intensity during energy generation, for
instance, Australia and China. Moreover, regulations form another driving force to
push manufacturers to practice energy efficiency measures more actively. The car-
bon tax passed in Australia is a good example, which has already raised the aware-
ness among its industries. In addition to government action, customers are also
starting to demand more environmental-friendly products. Thus, improved energy
and eco-efficiency would not only enhance the competitiveness of the manufactur-
ers, but also become a key survival skill in future markets.
Chapter 1
Motivation Research Question
Introduction
Research Objectives
Chapter 2
What is the What is the
A Critical energy efficiency? eco-efficiency ?
Review
Characterization of Environmental
Energy efficiency impacts
Characterization of Eco-efficiency
Chapter 4 Turning
Energy
Milling
efficiency of
selected Grinding
manufacturing Injection Molding
processes
Electrical Discharge Machining (EDM)
Efficiency
Modified Approach for Strategies
industrial applications
References
International Energy Agency (IEA) (2008) Worldwide trends in energy use and efficiency-key
insights from IEA indicator analysis, Online publication: http://www.iea.org/Papers/2008/
Indicators_2008.pdf, Last visited 01/02/2012
UNFPA (2011) State of world population 2011, online publication: http://www.unfpa.org/swp/,
Last visit 01/03/2012
Chapter 2
Review on the State of the Research
in Energy and Eco-efficiency
of Manufacturing Processes
Energy efficiency now has a frequent appearance in the public domain, which has
been discussed intensively in relation to economics, industrial competitiveness,
energy security, as well as global warming and climate change (Eichhammer and
Mannsbart 1997). In general, energy efficiency refers to the ratio of the useful out-
put of a process to the energy input into a process. Better energy efficiency means
using less energy to produce the same amount of useful output. Although it is
primarily an engineering term, the definition has already extended to encompass
2.1.1Thermodynamic Approach
The thermodynamic indicators are often defined as the science of energy and
energy processes, where the heat content or work potentials are measured in terms
of thermodynamic terms, such as enthalpy, entropy, exergy and so forth (Patterson
1996). These indicators offer unique and objective measures for a given process
in the context of physical conditions, i.e. temperature, pressure, etc. From an
enthalpy point of view, the energy efficiency () can be defined as Eq.2.1.
Hout
H = (2.1)
Hin
where Hout refers to the sum of useful energy outputs of a process; Hin refers to
the sum of all the of the energy inputs into a process.
In the context of industries and manufacturing, the useful energy is generally
associated with the minimum energy requirements to perform a task. For instance,
Giacone and Manc (2012) took a glass melting case to demonstrate the calcu-
lation procedure of thermodynamic energy efficiency. The useful energy output
was computed as the sum of theoretical heat required regarding the temperature
changes and chemical reactions of glass melting. However, this thermodynamic
approach was only validated on the thermal processes or a furnace. Other machin-
ing processes, such as turning, milling and grinding, were initially excluded. For
those processes, it is difficult to determine the minimum energy requirement
from a thermodynamics perspective, since the temperature changes in the chip-
formation areas are extremely sensitive to the combination of workpiece material,
tool geometry, cutting parameters and other process conditions (OSullivan and
2.1 What Is Energy Efficiency 7
Cotterell 2001). One alternative is to use cutting forces to estimate the minimal
energy requirement of a machining process, as shown in Eqs.2.2 and 2.3.
Pt = Fc V (2.2)
t2
E= Pt dt (2.3)
t1
where Pt refers to the instantaneous power; Fc refers to the cutting force; V refers
to the cutting speed; E refers to the energy consumption from time t1 to t2.
There were different theories and approaches for cutting force prediction, such
as orthogonal machining theory, empirical modelling, finite element method,
etc. As cutting force prediction is not the main research objective, a few leading
researches about this topic have been reviewed.
Oxleys model is one of the widely accepted theories of cutting force predic-
tion. He has investigated the mechanics of machining based on a geometrical
model of chip formation (Oxley 1998). This model is very useful for predict-
ing cutting forces, tool life as well as optimizing cutting conditions. However,
this model still had several major limitations. One of them is that the model was
limited to plain carbon steel work materials. To extend this approach, as Oxley
suggested, the high strain-rate/high temperature flow stress of other materials in
appropriate machining conditions should be developed. Another limitation of this
model is that it assumed a plane cutting face tool, whereas many industrial tools
have chip breaking devices such as obstructions located on the tools face. Also,
the model assumed that the cutting edge keeps perfectly sharp during the machin-
ing process. Lee has extended Oxleys machining theory in 2007 (Lee 2007).
The author conducted a series of experiments to verify the methodology under
various cutting conditions such as cutting speed, feed, tool nose radius, material
hardness etc. Lee introduced the magnitude of tool radius into this model, which
improved the model by not assuming that the cutting edge is perfectly sharp.
Johnson-Cooks flow stress model also has been included in the model in order
to extend the range of Oxleys model. However, Lee only experimentally vali-
dated this approach on hardened alloy steel. In 2009, Liu etal. presented a sim-
ilar work, who also introduced Johnson-Cooks flow stress model into Oxleys
model (Liu etal. 2009). They validated the methodology only on aluminum alloy
work pieces.
Armarego is another leading researcher in the field of cutting force modelling
for turning and milling processes. Different cutting force models have been pub-
lished through the use of both theoretical and empirical approach. The theoreti-
cal investigation was initially conducted to analyze two oblique cutting processes
(Armarego and Wiriyacosol 1978). The model has suggested the existence of a
common oblique cutting theory, which has formed the fundamental tool for further
researches on cutting force modelling. Different models have been published for
turning and milling processes (Whitfield and Armarego 1986; Budak etal. 1996;
Armarego and Samaranayake 1999). Although the cutting force models showed
8 2 Review on the State of the Research in Energy
a high accuracy of predicting cutting forces, the models were only validated for
a specific case. Critically, the consistency among those cutting force models
was lacking. The empirical modelling approach was also used in some of the
researches (Armarego etal. 2000). In that paper, a large amount of available data-
base and empirical models has been reviewed since Taylors work in 1907. Some
attempts to optimize the cutting force, surface roughness and tool life have been
documented in Armarego and Browns (1969) seminal book. Although the existing
equations have covered a wide range of process conditions, the values of expo-
nents and constants in the empirical equations were found unreliable.
Other researchers have also provided different types of cutting force models
(Kline etal. 1982; Stephensen 1989; Axinte 2001). The models have all claimed
achieving a great accuracy, but the uncertainty of the cutting forces has been also
stated.
For a more complex machining process, like grinding, the cutting force pre-
diction is more problematic. For instance, grinding force consists of three stages,
ploughing, cutting and rubbing. Each one of them has resulted in a complicated
model (Li and Fu 1980; Liu etal. 2008; Doman etal. 2009; Durgumahanti etal.
2010). Thus the estimations of minimum energy requirement based on cutting
forces are not as applicable as they may first appear.
Another alternative for estimating the minimal energy consumption is to use the
specific cutting energy or specific grinding energy for material removal processes.
Kalpakjian and Schmid have included a table of specific cutting energy for different
materials in their book (Kalpakjian and Schmid 2005). The recommended value for
steel ranged from 2.7 to 9.3ws/mm3. However, how to determine the specific
energy was not presented. Malkin and Joseph (1975) presented a more detailed
work for grinding process, plotting specific grinding energy against the specific
material removal rate. However, the exponent and constant were still missing for
an applicable estimation. He also suggested an average enthalpy increase between
ambient temperature and liquid state as 10.5kJ/cm3 for iron and steels (Malkin and
Guo 2008). Despite the approximation, the value tested was over-estimated since
the material does not all melt completely. Besides the academic researches, the tool
suppliers also provide some useful information for specific cutting force estimation.
SECO tools, a major turning and milling insert supplier, has included a specific
cutting force (kc) calculation in its catalogue, as Eq.2.4 (SECO 2009).
1 0.010
kc = kc1.1 (2.4)
(f sin )mc
where f refers to the feed rete; refers to the approach angle; 0 refers to the rake
angle; mc refers to material exponent; kc1.1 refers to the shear stress of the work-
piece material. A complete dataset of exponents and constants for different cutting
tools and workpiece material can be found in the catalogue, which makes the spe-
cific cutting energy estimation most applicable. However, it can be only applied to
turning and milling processes.
Despite the sophisticated calculation of minimal energy consumption, the defi-
nition of useful energy output has been criticized by different authors due to the
2.1 What Is Energy Efficiency 9
implicit value judgment (Boulding 1981; Patterson 1996). For instance, the turn-
ing process also requires energy for spindle rotation; if only the cutting energy is
considered as useful output, all the other energy requirements are considered as
waste heat, resulting in a biased estimation of energy efficiency.
In addition, another problem with the enthalpy energy efficiency is that it does
not consider the energy quality of the inputs and the useful outputs (Patterson
1996). However, this problem becomes significant when applying to a complex
system where a mix of energy resources is used. For instance, the majority of
manufacturing processes require electricity as energy input, but the electricity gen-
eration and the material processing consume various natural resources. To over-
come the problem, the thermodynamic quality measures are introduced, such as
exergy. The definition of exergy is, the amount of work obtainable when some
matter is brought to a state of thermodynamic equilibrium with the common com-
ponents of the natural surroundings by means of reversible processes involving
interaction only with the above mentioned components of nature (Szargut etal.
1988). Gutowski etal. has used this concept to provide a platform to evaluate
the environmental impacts among material preparation phases to manufactur-
ing process, as shown in Fig.2.1. The methodology was adapted from several
thermodynamics books (Szargut etal. 1988; Smith etal. 2001; Gyftopoulos and
Beretta 2005). As the Fig.2.1 shows, there are three important aspects of manu-
facturing processes: (1) the energy requirements for the materials; (2) the energy
requirements for manufacturing processes themselves; and (3) the efficiency of
the material and exergy transformations in manufacturing processes. The defi-
nition of energy, enthalpy, entropy, heat, work, temperature and exergy can be
Heat Temperature
Work in
Work out
Heat Temperature
2.1.2Physical-Thermodynamic Approach
Besides the challenges of determining the useful output in terms of either heat
content or work potential, the thermodynamic measurement does not reflect the
end use service required by consumers. Hence, the efficiency ratios measure the
output in physical units rather than in thermodynamic terms, whereas the input
energy is still measured in traditional thermal terms, such as joule. For instance,
the function of a general turning process is to remove material to achieve a round
profile. This output can therefore be measured by cm3.
As suggested by Patterson, one advantage of using physical measures is
that they can be objectively measured as thermodynamic measures can, mean-
while they have the added advantage that they directly reflect what consumers
are actually requiring in terms of an end use service. Furthermore, the mar-
ket value of the output can be further converted from the physical measures,
which enables longitudinal analyses (Patterson 1996). Therefore, these hybrid
physical-
thermodynamic measures of energy are widely used in industrial,
residential, commercial and other sectors.
Obviously, the energy intensity indicator, which is the input energy consump-
tion per part or a unit service, is the inverse of the physical-thermodynamic meas-
ures of energy efficiency. It is also called as specific energy consumption, SEC, as
shown in Eq.2.5. The use of energy intensity or SEC can be found in numerous
industrial cases. For instance, Phylipsen etal. defined the energy efficiency indica-
tors for the iron and steel, aluminum, cement, and other energy intensive indus-
tries. The specific energy consumption was selected to compare among different
countries and industries (Phylipsen etal. 1997). Tanaka compared different energy
efficiency measurements, such as absolute energy consumption, energy intensity
and thermal efficiency. The specific energy consumption was finally selected for
the case of iron and steel industry in Japan (Tanaka 2008).
Energy input into a process
SEC = (2.5)
Physical output of a process
For a unit manufacturing process, the specific energy consumption is more favora-
ble than other energy efficiency indicators. The aforementioned specific cutting
energy or specific grinding energy is actually the form of energy intensity meas-
ures. Notably, these indicators only refer to the minimum energy requirement,
2.1 What Is Energy Efficiency 11
which is completely different from the specific energy consumption of a unit pro-
cess. In other words, the energy consumptions of auxiliary components need to
be taken into account. In that sense, the total energy consumption of a machine
tool needs to be measured in addition to the cutting energy or the spindle energy
demand.
More importantly, the specific energy consumption is not constant for a unit
manufacturing process. Owing to its dynamic nature, the specific energy con-
sumption varies according to different process parameters, workpiece materials.
For example, Eq.2.4 has already shown the dynamics of specific cutting force,
which directly affects the loads on the machine spindle as well as the total energy
input into the process. Therefore, a model is required to characterize the relation-
ship between the specific energy consumptions and process parameters. Ideally,
the model should offer a reliable prediction of specific energy consumption for a
given process, which will enable further environmental analysis and development
of energy efficiency strategies.
Prior to this research, the prediction of SEC for unit manufacturing process
remains unreliable or inapplicable. The existing models or methodologies fail to
cope with the total energy consumption of a machine tool, or the dynamic behav-
iors of the energy consumption, as reviewed below.
One attempt is to link the minimal cutting energy requirement with the total
energy consumption of a machine tool, which primarily assumes that the mini-
mal cutting energy is predictable. In other words, the machine or motor efficiency
() is targeted for modelling. However, no existing work has been conducted at
machine tool level. Draganescu etal. (2003) attempted to model the spindle motor
efficiency on a vertical milling machine. The motor efficiency has been defined as
Eq.2.6, where Pc refers to the minimal cutting power; Pmc refers to the consumed
power by the spindle drive motor.
Pc
= (2.6)
Pmc
The authors used the response surface methodology to establish the relationship
between the machine tool efficiency and working parameters such as spindle speed
(n) and torque (Mt). The derived model for the tested milling machine is shown in
Eq.2.7; and the efficiency surface response is shown in Fig.2.2.
= exp 9.136 + 2.362ln n + 1.135 ln Mt 0.166(ln n)2
0.141(ln Mt )2
0.083(ln n)(ln Mt )] (2.7)
12 2 Review on the State of the Research in Energy
Fig. 2.2The efficiency surface response for a tested spindle motor of a machine tool
(Draganescu etal. 2003)
As the model suggests, the efficiency of the spindle motor has shown a complex
trend against the load on the machine tool. However, the derived empirical model
for motor efficiency is unlikely to be applied for specific energy consumption of
the machine tool. Moreover, this statistical analysis was limited to one specific
machine when operating on aluminum alloy. The reliability of the efficiency
model cannot be guaranteed with other machines or processes.
On the contrary, disregarding the machine efficiency would result in failure
of the prediction of energy consumption. Klocke etal. (2010) has presented a
theoretical calculation of the cutting power. The total energy consumption of the
machine tool was assumed as the sum of idle power and cutting power. However,
the prediction has failed to match the measured energy consumption. The rea-
son was mainly due to the exclusion of waste energy (e.g. heat) during the metal
removing period. In other words, the energy consumption of a machine tool is not
just fixed power plus working power. The internal energy conversions and trans-
missions remains as a complex manner, yet unknown.
Other researchers consider the dynamics of machine tools as the different states
from power on to off, such as start-up, stand-by, ramp-up, processing, ramp-down,
and power-off. Dietmair and Verl (2008) accomplished a case study of energy
consumption forecasting based on the measurement of a machine tool at different
states. Figure2.3 shows the machine states over a measuring cycle.
2.1 What Is Energy Efficiency 13
Fig.2.3Operational states
and the measuring steps
(Dietmair and Verl 2008)
Fig.2.4Specific energy
requirements for various
manufacturing processes
(Gutowski etal. 2005)
2.1 What Is Energy Efficiency 15
The energy efficiency is not just an indicator. The definition given by World
Energy Council (WEC) suggests that Energy efficiency improvements refer to a
reduction in the energy used for a given service (heating, lighting, etc.) or level of
activity (WEC 2008). The understanding of energy efficiency has evolved from
a simple input/output ratio towards the global efforts for energy reduction. The
researches and practices in the field of manufacturing have been conducted in the
different levels from component, to unit process, and factory level.
At the component level, the researches of machine tools have also showed
great efforts to improve the energy efficiency. Abele etal. (2010) has reviewed
the state-of-the-art technologies and researches for machine tool main spindle
units. Normally, the spindle unit is the biggest energy consumer in the machine
tool. Different models and simulation tools have been developed for improv-
ing the design of spindle motor. However, the focus still remained on the quality
performance of the spindle unit, such as accuracy and reliability. There also exist
plenty of mechanical design solutions and drive methods to achieve a high-speed
production but not compromising any other quality indicators. Mori etal. (2011)
has given a perspective from machine tool builder side. A new acceleration control
method was developed to reduce the time for non-value adding activities, such as
tool change, positioning, acceleration, returning, etc.
At the unit process level, Munoz and Sheng (1995) has proposed an analyti-
cal approach to evaluate the environmental impact of machining processes by con-
sidering process mechanics, wear characteristics and lubricant flows. Although the
tool wear and the efficiency of applying cutting fluid can be improved, the gap
towards environmentally-conscious manufacturing remains large due to the exclu-
sion of energy consumptions. In addition, isolating the process from the machine
tool cannot provide the true efficiency of any manufacturing processes. Anderberg
etal. (2009) has developed a cost model to evaluate both cost and energy effi-
ciency of a CNC lathe machine. The proposed model covered multiple types of
cost, such as energy cost, machine cost, tool cost, operational cost, carbon dioxide
emission cost, etc. Although the direct energy cost was not comparable with direct
machining cost (tool cost and labor cost), the high energy consumption did reflect
16 2 Review on the State of the Research in Energy
the least cost-efficient alternative. The author also suggested a better knowledge
about the relationship between important machining parameters and energy con-
sumption will increase the energy efficiency for the CNC machining processes.
At the factory level, simulation has been used to characterize the energy and mate-
rial flow within the factory. Thiede and Herrmann (2010) presented a holistic view of
simulating a manufacturing system considering the interrelationship among processes
and the technical building services. But the dynamic behavior of individual process
was not modelled. The energy efficiency improvements would be more useful if the
unit energy consumption model can be integrated within the simulation system.
Besides the above mentioned energy efficiency practices, Herrmann etal.
(2009) provided an extended perspective. They have strategically pointed out
the limitation of solely improving energy efficiency. According to their sugges-
tions, all the relevant input and output flows should be considered, including heat,
compressed air, coolant, periphery system, etc. In other words, all the consumed
energy and resources need to be considered to avoid shifting the problem shifting
as well as to enable the overall efficiency of a machine tool. This suggestion leads
to the discussion of the second keyword of this research, eco-efficiency.
2.2What Is Eco-efficiency?
Thus, the concept of eco-efficiency should be applicable for unit processes and
transferable among different processes to magnify the benefits at corporate level or
throughout product life cycle.
Gutowski (2010) has initially discussed eco-efficiency for unit manufacturing
process, where he recommended that eco-efficiency is often the reciprocal of some
intensity metric, e.g. energy intensity. It is true for those processes whose energy
consumption dominates their environmental impacts. However, this perspective
should be extended by considering processes and materials which are needed to
support the actual value creation process, because these supporting materials (e.g.
coolant) and supporting processes (e.g. coolant filter) have significant environmen-
tal impacts. Therefore, a more holistic view of evaluating eco-efficiency for unit
process is necessary, but still remains absent.
One of the differences between energy efficiency and eco-efficiency is the con-
sideration of environmental impacts of the unit process. A series of environmental
analysis has been conducted among different manufacturing processes. Since the
environmental impact of the most tested process is mainly due to the electricity
usage, those environmental analyses all started with energy consumption studies.
Essentially, there is no difference between energy efficiency and eco-efficiency at
this point.
Kordonowy (2002) experimentally measured the energy consumption of selected
machines for his B.Sc. thesis. In his thesis, the energy consumed by a machine
could be broken down to three stages, such as constant start up stage, constant
operation stage, machining stage. A similar state-based measurement was con-
ducted to assign energy consumptions for each machine component. Kordonowy
further measured the machining power when applying different material removal
rates to the machines. According to the results, the machining power consumption
varies dramatically due to different material removal rates. But Kordonowy did not
explain how the process parameters impact on the energy consumption. Instead of
establishing the relationship between energy consumption and process parameters,
the research focused on the constant parts, and tried to theoretically calculate the
idle power consumption by using the machine specifications. However, this esti-
mation was unsuccessful especially when applied to the lathe machine (calculated
21810W vs. measured 1770W).
Nevertheless, Kordonowys thesis had provided some useful information for
further rough analysis. By referring Kordonowys findings, Dahmus and Gutowski
(2004) then presented a system level environmental analysis of machining pro-
cesses where grinding process was excluded. In this paper, the authors stated that
the energy necessary to actually cut the material is only an insignificant fraction of
the total energy consumption, and the differences between different cutting condi-
tions were ignored when attempting to assess the total system energy requirement.
18 2 Review on the State of the Research in Energy
Those statements were based on the previous research of Toyota production pro-
cesses by Gutowski etal. (2005). By creating an annual production scenario, the
actually machining energy consumption was assumed to be only 14.8% of the
total energy consumption. However, this analysis was embodied in the entire pro-
duction line for a long period of time. The fraction of machining energy consump-
tion was dependent on how many vehicles were produced during that time. Thus,
it is unacceptable to ignore the impact of machining parameters on total energy
consumption in a smaller scale of production or in unit process analysis. Later,
the roughly estimated energy consumption for material removal has then been
compared with other energy requirement such as material production, cutting fluid
preparation, tool preparation, tool construction and others. The results showed that
the embodied energy in the material dominate the energy involved in the material
removal processes. However, the energy requirement during the material produc-
tion does not belong to the unit process, as the material is not created or consumed
by the process itself. Therefore, the involvement of embodied energy of raw mate-
rial can result in a biased conclusion of environmental impacts of unit process.
In 2006, Thiriez and Gutowski conducted series of power measurements
on three types of injection molding machines: hydraulic, hybrid and all-electric
machines. A similar procedure was conducted to allocate energy consumption
to each component. In addition, Thiriez (2006) attempted to model the relation-
ship between energy consumption and throughput rate in his thesis. However,
the results were quite poor. The derived regression models showed low R-square
values for the tested machine tools (less than 0.5); and, the models did not agree
with the proposed exergy framework (Eqs.2.8 and 2.9).
In that research, besides the basic energy consumption analysis, the life cycle
inventory analysis of injection molding process was also included. It has identi-
fied that the major contributor for the environmental impacts is due to raw mate-
rial production. Similar to the results of material removal processes, the embodied
energy in the raw material has overweighed the energy consumption during other
stages. But this does not mean that the energy consumption during manufacturing
stage can be neglected. On the contrary, the manufacturing stage is more dynamic
than material production; and the total energy consumption due to manufacturing
is still considerable. Hence, the improvement of energy consumption and environ-
mental impacts of unit manufacturing processes is still demanded.
Overall, the above mentioned environmental analysis only considered an average
specific energy consumption of a unit process, which resulted in constant results of
environmental impact of tested process. Consequently, the process dynamics have
been excluded in the current studies. More importantly, the energy for material pro-
duction was incorrectly included in the analysis of unit process. From the unit pro-
cess point of view, the material is only changed in its geometric, surface and other
features from raw material to end product. The material amount is not created from
zero to certain gram for an injection molding process; or, the chips do not disappear
for a turning or milling processes. Therefore, the embodied energy for a material
should be considered from a product point of view, not unit process; and the envi-
ronmental impacts during material production stage need to be studied separately.
2.2 What Is Eco-efficiency? 19
2.2.3LCC/LCA Approach
Calculating eco-efficiency based on a ratio between life cycle costs and life cycle
assessment (LCC/LCA) is proposed by different authors (Kicherer etal. 2007;
Huppes and Ishikawa 2005; Lyrstedt 2005).
According to Westkmper etal., LCA is a technique to assess environmental
impacts associated with a function. Generally, it is applied to a product, but the
product is defined as an object to fulfil a certain function. A holistic view is essen-
tial for LCA, which should cover the entire life cycle of the product or the func-
tion, as well as different types of environmental impacts. The LCC refers to the
valuation of the costs of production, installation, usage and disposal (Westkmper
etal. 2000). By assuming that life cycle cost reflects the value of the product or
a function, the LCC/LCA can theoretically provide the information for its eco-
efficiency (Lyrstedt 2005; Kicherer etal. 2007). However, the proposed cases were
originally for the assessment of a product or company in order to support decision
making. Huppes and Ishikawa (2005) provided a theoretical approach for assess-
ing eco-efficiency of society. To adapt this LCC/LCA method for unit process, the
information about every involved life cycle activities need to be provided. Taking
grinding process as an example, besides the electricity consumption, the resources
consumptions such as coolant, the grinding wheel also should be included in the
eco-efficiency analysis. In that sense, the cost and the environmental impacts of
producing and consuming those resources need to be estimated. However, none of
the information can be found in either previous literature or Life Cycle Inventory
(LCI) databases. In order to gather above information, it needs a joint effort from
machine tool builders, cutting tool suppliers, coolant makers and the users.
In short, due to the high demand of input information, the LCC/LCA is inap-
plicable for the case of unit process at the moment. Therefore, the methodology of
evaluating the eco-efficiency of unit process is required in this research.
This chapter has mainly reviewed the existing methods for characterizing energy
and eco-efficiency of unit manufacturing processes. The indicator for evaluat-
ing energy efficiency has been selected, and the specific energy consumption is
favored due to the objective measurements, as well as the meaningful reflection
of customer requirements. However, the current methodology faces shortages to
either describe or predict the dynamic behavior of unit manufacturing processes.
Therefore, there is an essential need to develop unit energy consumption models
to characterize the relationship between specific energy consumption and process
parameters, in order to evaluate the energy efficiency of a unit process.
The eco-efficiency for unit process is also under development. Although the
eco-efficiency can be simplified as the reciprocal of energy intensity for processes
that only consume electricity, other resource consumptions need to be taken into
20 2 Review on the State of the Research in Energy
account for a relatively complex process, such as grinding. In addition, the pro-
cess value remains undefined for the case of unit process. Therefore, the meth-
odology of characterizing eco-efficiency for unit process also needs a systematic
development.
References
Abele E, Altintas Y, Brecher C (2010) Machine tool spindle units. CIRP Annals Manufact
Technol 59(2):781802
Anderberg S, Kara S, Beno T (2009) Impact of energy efficiency on computer numerically con-
trolled machining. Proc Inst Mech Eng Part B J Eng Manufact 224(4/2010):531541
Aoe T (2007) Eco-efficiency and ecodesign in electrical and electronic products. J Clean Prod
15:14061414
Armarego EJA, Brown RH (1969) The machining of metals. PrenticeHall Inc., New Jersey
Armarego EJA, Samaranayake P, (1999) Performance prediction models for turning with
rounded corner plane faced lathe tools. II. Verification of models. J Mach Technol
3(2):173200
Armarego EJA, Wiriyacosol S (1978) Oblique machining with triangular form toolsI.
Theoretical investigation. Int J Mach Tool Design Res 18(2):6780
Armarego EJA, Ostafiev D, Wong SWY, Verezub S (2000) Appraisal of empirical modeling and
proprietary software databases for performance prediction of machining operations. J Mach
Sci Technol 4(3):479510
Axinte DA, Belluco W, De Chiffre L (2001) Evaluation of cutting force uncertainty components
in turning. Int J Mach Tools Manufact 41(5):719730
Boulding KE (1981) Evolutionary economics. Sage Publications, CA
Budak E, Altintas Y, Armarego EJA (1996) Prediction of milling force coefficients from orthogo-
nal cutting data. J Manufact Sci Eng Trans ASME 118(2):216224
Dahmus JB, Gutowski T (2004) An environmental analysis of machining. in American Society of
Mechanical Engineers. Manufacturing Engineering Division, MED. Anaheim
Dietmair A, Verl A (2008) Energy consumption modeling and optimization for production machines.
In: IEEE international conference on sustainable energy technologies, ICSET, Singapore
Doman DA, Warkentin A, Bauer R, (2009) Finite element modeling approaches in grinding. Int J
Mach Tools Manufact 49(2):109116
Draganescu F, Gheorghe M, Doicin CV (2003) Models of machine tool efficiency and specific
consumed energy. J Mater Process Technol 141(1):915
Duflou JR, Kellens K, Dewulf W (2010) Unit process impact assessment for discrete part manu-
facturing: a state of the art. CIRP J Manufact Sci Technol (in press)
Durgumahanti USP, Singh V, Rao PV (2010) A new model for grinding force prediction and
analysis. Int J Mach Tools Manuf 50(3):231240
Eichhammer W, Mannsbart W (1997) Industrial energy efficiency-indicators for a European
cross-country comparison of energy efficiency in the manufacturing industry. Energy Policy
25(79):759772
Giacone E, Manc S (2012) Energy efficiency measurment in industrial processes. Energy
38(1):331345
Gutowski T (2010) The efficiency and eco-efficiency of manufacturing. Int J Nanomanuf
6(14):3845
Gutowski T, Murphy C, Allen D, Bauer D etal (2005) Environmentally benign manufacturing:
observations from Japan, Europe and the United States. J Clean Prod 13(1):117
Gutowski T, Dahmus J, Thiriez A, Branham M, Jones A (2007) A thermodynamic characteriza-
tion of manufacturing processes. In: IEEE International symposium on electronics and the
environment, Orlando
References 21
Gutowski T, Branham MS, Dahmus JB, Jones AJ, Thiriez A, Sekulic DP (2009) Thermodynamic
analysis of resources used in manufacturing processes. Environ Sci Technol 43(5):15841590
Gyftopoulos EP, Beretta GP (2005) Thermodynamics: foundations and applications. Dover
Publications Inc, New York
Herrmann C, Thied S, Zein A, Ihlenfeldt S, Blau P (2009) Energy efficiency of machine tools:
extending the perspective. In: 42nd CIRP international conference on manufacturing systems.
Grenoble
Huppes G, Ishikawa M (2005) A framework for quantified eco-efficiency analysis. J Ind Ecol
9(4):2541
Kalpakjian S, Schmid SR (2005) Manufacturing engineering and technology, 5th edn. Prentice
Hall
Kicherer A, Schaltegger S, Tschochohei H, Pozo BF (2007) Eco-efficiency combining life cycle
assessment and life cycle costs via normalization. Int J Life Cycle Assess 12(7):537543
Kline WA, DeVor RE, Lindberg JR (1982) The prediction of cutting forces in end milling with
application to cornering cuts. Int J Mach Tool Design Res 22(1):722
Klocke F, Schlosser R, Lung D (2010) Energy and resource consumption of cutting processes
how process parameter variation can optimise the total process efficiency. In: 17th CIRP
international conference on life cycle engineering. Hefei
Kordonowy DN (2002) A power assessment of machining tools, in mechanical engineering.
Massachusetts Institute of Technology, Cambridge
Lee TH (2007) An experimental and theoretical investigation for the machining of hardened alloy
steels. PhD thesis in School of mechanical and manufacturing engineering, UNSW, Sydney
Li L, Fu J (1980) A study of grinding force mathematical model. CIRP Annals Manufact Technol
29:245249
Liu Q, Chen X, Wang Y, Gindy N (2008) Empirical modelling of grinding force based on multi-
variate analysis. J Mater Process Technol 203(13):420430
Liu HT, Sun YZ, Lu ZS, Han LL (2009) The prediction model of cutting forces based on
Johnson-Cooks flow stress model. Key Eng Mater 392394:16
Lyrstedt F (2005) Measuring eco-efficiency by a lcc/lca ratio an evaluation of its applicabilitya
case study at ABB. Master Thesis, Chalmers University of Technology, Gteborg
Malkin S, Guo C (2008) Grinding technologytheory and applications of machining with abra-
sives, 2nd edn. Industrial Press, New York
Malkin S, Joseph N (1975) Minimum energy in abrasive processes. Wear 32(1):1523
Mori M, Fujishima M, Inamasu Y, Oda Y, (2011) A study on energy efficiency improvement for
machine tools. CIRP Annals Manufact Technol 60(1):145148
Munoz AA, Sheng P (1995) An analytical approach for determining the environmental impact of
machining processes. J Mater Process Technol 53(34):736758
OSullivan DO, Cotterell M (2001) Temperature measurement in single point turning. J Mater
Process Technol 118:301308
Oxley PLB (1998) Development and application of a predictive machining theory. J Mach Sci
Technol 2(2):165189
Patterson MG (1996) What is energy efficiency? Concepts indicators and methodological issues.
Energy Policy 24(5):377390
Phylipsen GJM, Blok K, Worrell E (1997) International comparisons of energy efficiency-meth-
odologies for the maufacturing industry. Energy Policy 25(79):715725
Renaldi Kellens K, Dewulf W, Duflou JR, (2011) Exergy efficiency defnitions for manufacturing
processes. In: 18th CIRP international conference on life cycle engineering. Braunschweig
SECO Tools (2009) MN2009-turning. The Machining Navigator Catalogue
Sinkin C, Wright CJ, Burnett RD (2008) Eco-efficiency and firm value. J Acc Public Policy
27:167176
Smith JM, Van Ness HC, Abbott MM (2001) Chemical engineering thermodynamics, 6th edn.
McGraw-Hill
Stephensen DA (1989) Material characterization for metal-cutting force modelling. J Eng Mater
Technol Trans ASME 111(2):210219
22 2 Review on the State of the Research in Energy
Szargut J, Morris DR, Steward FR (1988) Energy analysis of thermal, chemical, and
metallurgical processes. Hemisphere Publishing Corporation and Springer-Verlag, New York
Tanaka K (2008) Assessment of energy efficiency performance measures in industry and their
application for policy. Energy Policy 36(8):28872902
Thiede S, Herrmann C (2010) Simulation-based energy flow evaluation for sustainable manufac-
turing systems. In: 17th CIRP international conference on life cycle engineering. Hefei
Thiriez A (2006) An environmental analysis of injection molding. Master thesis in MIT,
Cambridge
Thiriez A, Gutowski T (2006) An environmental analysis of injection molding. In: IEEE interna-
tional symposium on electronics and the environment, Scottsdale
Westkmper E, Alting Arndt (2000) Life cycle management and assessment: approaches and
visions towards sustainable manufacturing (keynote paper). CIRP Annals Manufact Technol
49(2):501526
Whitfield RC, Armarego EJA (1986) Predictive milling force models for cad/cam systems. Aust,
Inst of Engineers, Newcastle
World Bussiness Council for Sustainable Development (WBCSD) (2000) Eco-efficiency: creat-
ing more value with less impact, Geneve
World Energy Council (2008) Energy efficiency polices around the world: review and evaluation,
online publication: http://www.worldenergy.org/publications/energy_efficiency_policies_
around_the_world_review_and_evaluation/default.asp, Last visited 1 March 2012
Chapter 3
Methodology for Characterizing Energy
and Eco-efficiency of Manufacturing
Processes
3.1.1Empirical Modelling
8000
6000
The area = SEC (kJ/cm or kJ/g)
Effective Power (W)
2000
Stand-
Stand-by
by
0
Duration of Processing 1 unit material Time (s)
both processing load and base load. It can also be transferable and comparable
among different machine tools and manufacturing processes (Kara and Li 2011).
Nevertheless, the unit of SEC varies depending on the feature of the processes. For
example, turning is a material removal processes which is configured by volumetric
parameters. Thus, SEC for turning process would be the total energy consumption
of a lathe to remove a unit volume of material. For the case of injection molding,
plastics are firstly melted in a heated chamber and then solidified in a mold, where
the processed material is normally measured by mass. Hence, SEC for an injec-
tion molding process would be the total energy consumption of the machine tool
to inject a unit mass of material. It should also be noted that the unit of SEC can be
easily converted from per volume to mass by multiplying the material density, or
vice versa. Therefore, the modelling objective can be interpreted as characterizing
the relationship between process parameters and specific energy consumptions.
X1 X2 0
T0 =
(3.1)
S12 S22
n1 + n2
2
S12 S22
n1 + n2
DF = 2 2 2 2 (3.2)
S1 /n1 S2 /n2
n1 1 + n2 1
where
T0 is the test statistic;
DF is the degrees of freedom, which should be rounded down to the nearest
integer;
X1, X2 are the means of two group data;
n1, n2 are the number of data;
S1, S2 are the standard deviations of two group data; and,
0 is the test of interest, in this case, it should equal to 0.
3.1 Methodology of Energy Efficiency 29
The statistical chart of t distribution can be used to determine the P-value. If the
P-value is greater than 0.05, the statistical analysis would suggest that the main
effects of the tested factor are insignificant, thus the factor will be grouped as a
held-constant factor for further steps. Otherwise, the factor will be considered as a
design factor, and included for model development.
After factor-screening, all the design factors are determined for model develop-
ment. However, the interaction among design factors may affect the reliability of
derived models. Hence, OFAT will no longer be suitable for modelling purpose.
As mentioned before, factorial experiments are favored by many researchers for
assessing the interactions between factors, but the typical 2-level factorial experi-
ment design (2k design) has its own drawbacks. It initially assumes that the regres-
sion between the factors and model response is almost linear. Therefore, a more
complicated approach, Response Surface Methodology (RSM), is then introduced.
Response Surface Methodology, or RSM, is a collection of mathematical and
statistical techniques useful for modelling and analysis in applications where a
response of interest is influenced by several variables; and, the original objective
is to optimize this response (Montgomery 2009). RSM is also a sequential proce-
dure. If the response is well modelled by a linear function of the independent vari-
ables, the approximating function is the first-order model (Eq.3.3).
Y = 0 + 1 x1 + 2 x2 + + k xk + (3.3)
If there is curvature in the system, a polynomial of higher degree must be used,
such as the second-order model (Eq.3.4).
k
k
Y = 0 + i xi + ii xi2 + ij xi xj + (3.4)
i=1 i=1 i<j
In theory, the RSM can provide the most reliable empirical models in this
research. But the necessity of second-order model remains uncertain at this stage.
Similar to factor-screening, there are multiple methods for experimental designs
of RSM, such as full 3k factorial design, Box-Behnken design, and central com-
posite design (NIST 2010). The first two methods are excluded in this research
due to the disadvantage of run size. Central Composite Design (CCD), how-
ever, is a favorable option of many engineering projects as well as this research.
It contains an embedded factorial or fractional factorial design with center points
which allow estimation of curvature (NIST 2010). There are three types of CCD:
Circumscribed Central Composite (CCC), Inscribed Central Composite (CCI), and
Face Centered Composite (CCF). The difference among those three specific meth-
ods is illustrated in Fig.3.3.
The pre-defined lower and upper levels of each factor are noted as start points.
The additional levels generated by CCD are the center points shown as red dots in
Fig.3.3. It clearly visualizes the center points derived by CCC exceed the range of
the start points, which may result in infeasible runs for a constraint system. On the
contrary, CCI can be seen as a scaled down CCC. It causes difficulties in selecting
starting points whilst using center points to test the system limits. Alternatively,
30 3 Methodology for Characterizing Energy
+1
-1
CCF only requires the maximum range of a factor as starting points, and generates
the center points as the midpoints between the maximum and minimum value, the
method of which is considered as the best choice for this research.
The statistics package, MiniTab Version 15, is used to generate the schedule
of experiments. It offers a design of experiment module with the function of RSM
and CCF. In this research, trial runs are required to determine the range of each
design factor. Experts opinion can be also helpful for this case. Once the ranges of
all design factors are determined, the upper and lower levels of each factor can be
input into MiniTab. In order to avoid any human induced trend, the randomiza-
tion function needs to be enabled when generating the experiment schedule.
In short, there are three key steps in stage I: firstly, all the involved factors are
classified into different groups; secondly, a factor-screening procedure is con-
ducted for uncertain factors by using one-factor-at-a-time method; then, response
surface methodology and face centered composite design is used to generate a full
schedule of experiments with the assists of MiniTab.
Collecting accurate and adequate data from the experiments is essential for this
research. Kara etal. (2011) pointed out the importance of energy metering and
monitoring in manufacturing systems in their keynote paper. They have provided
not only the guidelines for establishing an energy metering and monitoring sys-
tem, but also discussed the technical aspects from factory level to unit process
level. As the keynote suggested, a clear metering task needs to be defined in the
first place. Then the measurands, resolution and communication interface can be
selected accordingly.
The energy consumption behavior of a machine tool is highly dynamic in
most cases. For example, current motor technologies offer rapid acceleration
and deceleration in order to achieve operational readiness as quick as possible.
Consequently, instantaneous power peak occurs during these processes. This
behavior desires a high resolution of energy metering and monitoring system to
3.1 Methodology of Energy Efficiency 31
L1
L2
L3
N
Main
Bus
(b) (c)
Hardware Subsystem
(d) (e)
Monitoring Platform
(f) (g)
Fig.3.4Structure of energy metering and monitoring system for unit processes. a Current
clamp FLUKE i200s. b Voltage Module NI 9225. c Current Module NI 9229. d NI cDAQ-
9172. e Laptop PC. f Block diagram of LabVIEW Program. g Front panel of LabVIEW Program
3.1 Methodology of Energy Efficiency 33
The objective of this stage is to statistically express the relationship between the
process variables and observed responses. In order to achieve the reliability of the
empirical models, the statistical techniques should be coupled with good engi-
neering or process knowledge and common sense (Montgomery 2009).
As specific energy consumption (SEC) is not directly measurable, the proposed
responses need to be computed from other measurable values. Firstly, the total
energy consumption during the processing period is integrated from the recorded
data. Then, the total volume or mass of processed material is theoretically calcu-
lated or directly measured. Since the current computer numerical control (CNC)
machine tools have achieved a high level of precision, the volume of processed
material can be assumed equal to what each process program commands. For the
case of injection molding, the mass of injected material varies depending on
thematerial type, melting temperature, viscosity of melted material, etc. Thus, all
34 3 Methodology for Characterizing Energy
SSE
R2 = 1 (3.5)
SST
where, SSE refers to the error sum of squares (or unexplained variability); and, SST
refers to the total sum of squares (or total variability). Each of these can also be
calculated according to Eqs.3.6 and 3.7.
n n
2
e2i
SSE = yi yi = (3.6)
i=1 i=1
n
SST = (yi y)2 (3.7)
i=1
where, yi is the observed response; yi is the predictor value; ei is the error in the fit
of the model to the ith observation, or called residual; and, y is the observed mean
of response. If the R2 is larger than 0.9, the model can be concluded as sufficient,
as more than 90% of observed data can be explained with the model.
In addition, the residuals of a good model should be randomly distributed or
normal. Thus, the residuals need to be plotted against the predictor values. If the
distribution is not random or normal, it suggests that other factors have certain
impacts on the system which need to be statistically included in the model.
A more accurate model can be derived from multiple-linear regression. As any
regression model can be rewritten in a linear form regardless of the shape of the
observed surface, the derived models from curve estimation can be converted into
a linear format (Montgomery 2009). For instance, if the factor x1 fits the inverse
model for predicting response Y as in Eq.3.8; the reciprocal of x1 can be noted as
a new factor x1 ; whereby, the inverse model can be also written as the linear one
shown in Eq.3.9.
c1
Y = c0 + (3.8)
x1
1
Let x1 = x1 , then:
Y = c0 + c1 x1 (3.9)
By using the stepwise function of linear regression in SPSS, all the significant
factors are introduced to the model sequentially according to the significance of
their impacts. However, this method will certainly increase the complexity of the
model, and affect the consistency among different machines and processes. Thus,
the model derived from curve-estimation can be accepted as a generic model,
whereas the results of multiple-linear regression explore the detailed interrelation-
ship within the system.
36 3 Methodology for Characterizing Energy
This research aims to achieve the discrepancy between prediction and measure-
ment less than 10%. This index can be also used to compare the accuracy between
a simple model and relatively complex model derived from curve-estimation and
multiple-linear regression respectively. It would suggest whether there is a sig-
nificant benefit to introduce multiple factors in a model for predicting energy
consumption.
related?
Yes
Yes P-value <
Group as design factors
0.05
Conduct experiments
R2 0.9?
Yes
Residuals plots
No Multiple-linear
Random?
regressions
Yes
Validation runs
Stage IV
Difference No
10%?
Yes
Model accepted
Process Value
Ecoefficiency =
Process Environemental Impacts
= f (Process Parameters) (3.11)
Environmental impacts can be generally summarized into the following three cat-
egories according to the United Nations Environmental ProgramSociety for
Environmental Toxicology and Chemistry (UNFP-SETAC 2005):
40 3 Methodology for Characterizing Energy
Owing to a dynamic nature, process parameters play a central role linking both
process value and environmental impacts, as shown in Fig.3.8. The amount of
process output has a direct link with the process parameters, which determines
how much volume or weight of material can be processed during a unit time. The
process parameters have a more significant influence on other quality measure-
ments. The studies about surface integrity and process precision have generated a
number of models to characterize the relationship between process parameters and
quality measurements (Jawahir etal. 2011).
Section3.1 has proposed a methodology to derive SEC models for unit process.
The results of these SEC models can be directly used for evaluating the associated
environmental impacts. Thus, there exists a platform to encompass all the three
layers including process value, environmental impact and process parameters; and,
it is possible to quantitatively describe their interrelationships.
Based on above analysis, a methodology for evaluating eco-efficiency of unit
process is proposed as Fig.3.9. Due to the lack of information about the envi-
ronmental analysis of every energy and resource involved, the application requires
assumptions and simplifications.
3.3Summary
This chapter has provided methodologies for the evaluation of energy and eco-effi-
ciency of unit processes. According to the review in Chap.2, the specific energy
consumption (SEC) is the appropriate indicator for evaluating energy efficiency
of unit process. The proposed methodology uses empirical modelling to derive
3.3Summary 43
No
References
With the proposed methodology in Sect.3.1, this chapter selects a wide range of
manufacturing processes to characterize their energy efficiency. The presented
cases range from single cutting-edge machining (e.g. turning) to multiple cutting-
edges machining (e.g. milling), from geometrically defined cutting-edge machin-
ing to geometrically undefined cutting-edge machining (e.g. grinding), from metal
machining to thermoplastic processes (e.g. injection molding), from conventional
manufacturing processes to unconventional manufacturing processes (e.g. electri-
cal discharge machining). For each process, a number of different machine tools
are presented in order to validate the proposed methodology.
4.1.1Background
material, cutting parameters and other factors three motors are grouped as load
dependent components. However, a turret servo motor is only activated while chang-
ing the cutting tool. A hydraulic pump motor provides constant clamping pressure for
holding the workpiece, which requires relatively static power throughout the process.
Energy consumption of other components such as amplifiers, fans and other auxiliary
components remains constant and form the base load of the machine tool.
Five different CNC lathes have been selected for modelling (as displayed in
Fig. 4.2) and they cover a variety of machine ages, spindle power and transmis-
sion systems. The first two lathes were located in the manufacturing laboratory
of School of Mechanical and Manufacturing Engineering, the University of New
South Wales whilst the other three were in Sydney City TAFE.
4.1.2Design of Experiments
Since turning is a material removal process, in this case, specific energy consump-
tion (SEC) refers to the total energy consumption of a CNC lathe for removing 1cm3
material. The objective of the experiments is to derive the relationship between SEC
and process parameters. There are numerous factors that influence a cutting operation
in a turning process. These can be generally classified into one of five categories:
Tool condition: i.e. material, insert geometry, tool wear, etc.
Workpiece material: generally referring to hardness and machinability.
4.1 Energy Efficiency Characterization for Turning Processes 47
a b
c d
Cutting parameters: i.e. cutting speed (V), feed (f) and depth of cut (d).
Cutting fluids: whether or not a coolant is used, type of coolant, etc.
Operation environment: i.e. temperature, humidity, etc.
Cutting parameters determine the duration of removing 1cm3 material, thus they
were grouped as design factors. As different workpiece materials feature differ-
ent machinability, the available range of cutting parameters differs from material
to material. For instance, cutting aluminum is relatively a fast process, the cut-
ting speed of which has range from 200 to 400m/min; while the cutting speed for
machining low-carbon steel is normally under 150m/min (Kalpakjian and Schmid
2005). Therefore, the type of workpiece material was also considered as a design
factor.
Benefits of applying cutting fluid (wet machining) are to lubricate the cutting
zone, to avoid workpiece damage, as well as to extend the tool life. However, the
cost of coolant and the environmental concerns have driven the trend towards
dry machining (Sreejith and Ngoi 2000). Several materials are recommended
for dry machining, i.e. aluminum, copper, steels, etc. (Kalpakjian and Schmid
2005). From the energy consumption perspective, applying cutting fluid requires
additional energy to power the coolant pump. Since power requirement of cool-
ant pump remains constant, it can be added to the base load in the further stages.
Therefore, the type of cutting fluids was considered as a held-constant factor.
As discussed in Sect.2.1, cutting tools have significant impacts on cutting
forces, which indirectly affect the loads on both servo and spindle motors. The
type of cutting tool or insert is strongly dependent on workpiece material due to
different machinability. Hence, the insert type was coupled with the workpiece
material. According to Oxleys cutting force model, the geometric features of cut-
ting tool first need screening. Cutting rake angle was selected for factor-screening,
since different cutting force models universally suggest the significance of rake
angles impacts (Oxley 1998; Armarego and Wiriyacosol 1978; SECO 2009).
Rake angle tests have been conducted on the Colchester Tornado A50. 50mm
diameter mild steel bars were used for testing with constant cutting speed, feed
and depth of cut but various insert rake angles from 6 to 6. 100 samples taken
during the cutting process were randomly extracted from each run. Then, a t-test
was performed according to the research methodology proposed in Sect.3.1. The
results of t-test are listed in Table4.1. Since the P-value is larger than 0.05, the
rake angle is an insignificant factor for the total energy consumption of the CNC
lathe. The results can apply to other geometric factors of cutting tools, because
their impacts are only limited at tool tip and the power requirements at tool tip
only account for a small portion of the total power consumption of the machine
Nuisance Design
Factors Factors
Workpiece material
Tool wear
Cutting speed
Operation temperature
feed
Operation humidity
Depth of cut Specific
Energy
Consumption
Machine vibration Rake angle
Nose radius
Workpiece unit-to-unit
variation Cutting fluid
Nuisance Held-constant
Factors Factors
tool. Therefore, the geometric factors of cutting tools were assumed negligible and
determined as held-constant factors for this research.
Other factors such as tool wear and operation environment were uncontrolla-
ble in this case and were grouped as nuisance factors. An Ishikawa diagram was
developed to organize all the involved factors, as shown in Fig.4.3.
Three workpiece materials were selected for investigation, i.e. aluminum alloy,
mild steel and high tensile steel, as listed in Table4.2. Aluminum alloy has an
excellent machinability, whereas high tensile steel is relatively hard to machine.
Consequently, the available range of cutting parameters differs from material to
material. The levels of tested factors are listed in Table4.3. It should be noted
that the variable levels were modified according to machine capacity in order to
investigate the widest possible range of cutting parameters. Then the schedule of
experiments for each workpiece material was generated by MiniTab as discussed
in Sect.3.1.3. A sample schedule for testing aluminum alloy on the Colchester
Tornado A50 is presented in Appendix 1.
4.1.3Experiment Details
Tool suppliers offer a wide range of cutting tools in terms of shape, size, chip-
former, etc. The tools need to be carefully selected according to the workpiece
material and the desired surface finish. According to the factor-screening results,
geometric features of cutting tools were assumed negligible. Hence, it is both
expensive and unnecessary to test different insert types. In this research, the
50
cutting tool was selected for general machining purposes. Both insert holders and
inserts were supplied by ISCAR tool. The selected insert holder was MWLNR/L
2020K-06W and the selected insert was WNMG 06T308-PP with grade IC9025.
However, due to the difficulties of machining high tensile steel (4140), an alter-
native grade IC3028 was used. The PP chip-former offers very positive rake and
sharp edges, which is recommended for aluminum alloys and low carbon steels.
As the tool geometry was held constant, a number of tool damages occurred
when cutting 4140 with a high level of depth of cut and feed. The situation can
be improved by changing the chip-former type, which has a secure cutting edge.
Nevertheless, the impacts of tool wear need to be considered when investigating
the eco-efficiency of unit manufacturing processes, discussed later.
There were three stages of cutting within each experiment run, as illustrated
in Fig.4.4. An exemplary power measurement of one complete run was plotted
against time at 0.1s intervals as shown in Fig.4.4a. As the figure shows, the power
curve exactly mirrors each machining operation: first, the raw metal bar was cor-
rected to a universal size of 49mm in diameter, i.e. skin-off (see in Fig.4.4b, c);
second, two repetitive facing cuts were performed in order to examine spindle
acceleration, in which a constant surface speed was applied (see in Fig.4.4d);
third, a 50mm long horizontal cut was performed with the scheduled cutting
parameters and this was replicated an additional two times (see in Fig.4.4e).
The power curve showed a constant trend during horizontal cuts, where
the analysis of data obtained for the facing cuts became problematic due to the
inconsistency of MRR along single cuts. A spindle needs to accelerate exponen-
tially when the workpiece diameter decreases in order to create a constant MRR.
Thus, once the spindle reaches its capacity, 400 rev/min in the case of Colchester
Tornado A50, and is no longer able to compensate the reduction in the diameter
of the workpiece a decrease in MRR results. Moreover, facing operations only
account for a small portion of total removal work in practice, which is normally
less than 5% of the total removal volume. Therefore, the analysis was mainly
focused on the horizontal machining actions.
52 4 Energy Efficiency Characterization of Manufacturing Processes
(a) 12000
10000
Effective Power (W)
4000
2000
#1 #2 #3 #4 #5 #6 #7 Time
0
1 201 401 601 801 (0.1s)
(b) (c)
f f
=49mm
d 1mm
(d) (e)
f f
d 50mm
d
Fig.4.4Exemplary power curve of one turning experiment and illustrations of cutting steps.
a An exemplary power measurement of one turning experiment. b Skin-off facing cutcut #1.
c Skin-off facing cutcut #2. d Skin-off facing cutcut #3 and 4. e Skin-off facing cutcut #5,
6 and 7
4.1 Energy Efficiency Characterization for Turning Processes 53
4.1.4Regression Results
First, each of three horizontal cuts was separated as shown in Fig.4.4a. The
energy consumption of each cut (E) is the integration of the power over the cutting
period. The cutting volume (Q) was determined based on the scheduled cutting
parameters (Eq.4.1). Thus, SEC was also calculated (Eq.4.2) as the response for
modelling. Sample data including SEC values for the Colchester Tornado A50 are
shown in Appendix 2.
Q = r02 (r0 d)2 l (4.1)
E Pi dt
SEC = = (4.2)
Q Q
where r0 refers to the original radius of the workpiece; and, l refers to the length of
the horizontal cut, which was 50mm here.
Each design factor was then processed using curve-estimation in relation to
SEC. Although SPSS offers 11 different types of model, the R-square was very
low for using single cutting parameters to predict SEC. An example model plot
between cutting speed and SEC on the Colchester Tornado A50 is displayed in
Fig.4.5.
Cutting speed
(m/min)
Fig.4.5Model plot between cutting speed and SEC for the Colchester Tornado A50
54 4 Energy Efficiency Characterization of Manufacturing Processes
Table4.4Model summary of curve-estimation between MRR and SEC for the Colchester
Tornado A50
Model type Model summary Equation
R2 F-value DF1 DF2 P-value
Linear 0.295 74.003 1 177 0.000 y = 13.2 7.1x
Logarithmic 0.728 472.591 1 177 0.000 y = 2.3 7.52 ln x
Inverse 0.993 26,563.613 1 177 0.000 y = 1.5 + 2.2
x
Quadratic 0.498 87.325 2 176 0.000 y = 17.6 20.2x + 4.8x 2
Cubic 0.681 124.615 3 175 0.000 y = 24.9 55.6x + 38.2x 2 6.9x 3
Compound 0.661 345.720 1 177 0.000 y = 12.6 0.38x
Power 0.978 7732.117 1 177 0.000 y = 3.5x 0.79
S 0.761 563.737 1 177 0.000 ln y = 1.4 + 0.17
x
Growth 0.661 345.720 1 177 0.000 ln y = 2.5 0.97x
Exponential 0.661 345.720 1 177 0.000 ln y = ln 12.6 0.97x
Logistic 0.661 345.720 1 177 0.000
SEC (kJ/cm3)
MRR
(cm3/s)
Fig.4.6Model plot between MRR and SEC for the Colchester Tornado A50
Residuals (kJ/cm3)
Fig.4.7Residuals plot of the inverse model for the Colchester Tornado A50
56 4 Energy Efficiency Characterization of Manufacturing Processes
4.1.5Model Validation
4.2.1Background
Fadal VMC4020
Year: 1998
Axes No.: 3
Spindle No.: 1
Motor Power: 11 kW
axis of rotation perpendicular to the workpiece surface. All of the three machine
tools were located in the manufacturing laboratory of School of Mechanical and
Manufacturing Engineering, the University of New South Wales.
4.2.2Design of Experiments
Milling processes share a similar mechanism of chip formation with turning pro-
cesses. Instead of rotating the workpiece, the rotation of cutter generates cutting
forces for material removal. It is reasonable to assume that the SEC model for
milling processes would agree with that of turning, where the material removal
rate (MRR) plays a key role. Since the cutter contains multiple cutting edges, the
equation of calculating MRR for face milling is slightly different than the one for
turning, as shown in Eq.4.5.
MRR = w d v (4.5)
where w refers to the width of cut (mm); d refers to the depth of cut (mm); and, v
refers to the linear speed of the workpiece (mm/min) which is determined by the
selection of feed per tooth (f), number of teeth on cutter (n), rotation speed (N),
cutting speed (V) and the cutter diameter (D), as shown in Eq.4.6.
V
v =f nN =f n (4.6)
D
Design Design
Factors Factors
Cutter diameter Cutting speed
Workpiece material
Feed per tooth
Width of cut
Depth of cut
Wet or dry machining Specific
Energy
Tool wear Insert material Consumption
Operation environment Number of teeth
Machine vibration Insert shape
Workpiece unit-to-unit
Cutting fluid type
variation
Nuisance Held-constant
Factors Factors
4.2.3Experiment Details
The workpiece was pre-machined to a 50mm square bar with 200mm length (l).
Climb milling was performed, where the maximum chip thickness occurred at the
start of the cut. Each combination of cutting parameters was repeated twice with-
out coolant and a third time with coolant. A sample power curve and the schematic
illustrations of each tool path are illustrated in Fig.4.10 for the case of cutter
diameter 32mm.
4.2 Energy Efficiency Characterization for Milling Processes 61
(a) 25000
1st Dry 2nd Dry 3rd Wet
20000 Cut Cut Cut
Mori Seiki
Effective Power (W)
DuraVertical
15000 Path 3
Path1 5100
10000
Path 2
5000
0 Time
1 101 201 301 401 501 601 701 801 (0.1s)
(c)
(b) Tool Path
Path 1 15 mm
Path 3
d
v
(d)
10 mm
200 mm
N
(e)
25 mm
v 50 mm
Path 2
Fig.4.10Exemplary power curve of one milling experiment and illustrations of cutting steps
(cutter diameter 32mm). a Exemplary power curve of one complete experiment run (32 mm).
bSchematic illustration tool path. c Tool path 1. d Tool path 2. e Tool path 3
62 4 Energy Efficiency Characterization of Manufacturing Processes
Although the DECKEL MOHA DMU 60P (DMG DMU 60P) is a 5-axis
machining center, only three axes were used in this research. There were two main
reasons for this decision: one is to keep the consistency among all the three tested
milling machine tools; the second is that the motor size for the two additional axes
is much smaller than the others; correspondingly, the power demand of the spindle
drive motor and other auxiliary components outweighs the consumptions of the
other two servo motors significantly. Thus, the SEC model for DMG DMU 60P
can still be assumed to be representative and reliable.
First, each of three tool paths was separated as shown in Fig.4.10. The energy con-
sumption of each cut (E) is the integration of the power over the cutting period. The
cutting volume (Q) was determined for the scheduled cutting parameters (Eq.4.7).
Q=wdl (4.7)
Thus, SEC was calculated similarly according to Eq.4.2 as the response for
modelling.
Then, each design factor was processed with curve-estimation in relation
to SEC. The results are identical to those of the turning processes. Each single
cutting parameter resulted in very low R-square values; while, the MRR with an
inverse model can consistently achieve an R-square value of over 0.9 for all the
tested machine tools. The model plot and statistical results for the Mori Seiki Dura
Vertical 5100 are presented in Fig.4.11 and Table4.9 respectively.
The ANOVA analysis of inverse model and model plots for the other tested
machine tools can be found in Appendix 3. Although the current inverse model
has achieved a high R-square value, the residual analysis needs to be conducted
from a statistical point of view. Figure4.12 shows the residuals plotted against the
predicted SEC for the case of the Mori Seiki Dura Vertical 5100. There is a visible
trend between the residuals and the predicted value. Thus, stepwise multiple-linear
regression was performed afterwards.
Similar to the turning process, different machine tools showed different model
shapes as presented in Table4.10. There is a lack of consistency of the accepted
factors after final steps. More importantly, there was an insignificant growth
of R-square value from inverse model to the multiple-linear regression model.
For instance, the inverse model for the DMG DMU 60P has already achieved
an R-square value at 0.997; after adding 6 other factors, the R-square value only
increased by 0.001. Therefore, the contribution of the other factors for the SEC
model can be concluded negligible.
In fact, the inverse model is more practical than the multiple-linear regression
models, because it agrees with the model for turning processes as listed in Eq.4.4,
which enables cross comparison between turning and milling processes. The only
difference among the tested machine tools is the value of the coefficients c0 and
4.2 Energy Efficiency Characterization for Milling Processes 63
SEC (kJ/cm3)
MRR
(cm3/s)
Fig.4.11Model plot between MRR and SEC for the Mori Seiki Dura Vertical 5100
Table4.9Model summary of curve-estimation between MRR and SEC for the Mori Seiki Dura
Vertical 5100
Model type Model summary Equation
R2 F-value DF1 DF2 P-value
Linear 0.257 191.572 1 555 0.000 y = 12.9 5.7x
Logarithmic 0.730 1497.344 1 555 0.000 y = 2.2 6.562 ln x
Inverse 0.947 9844.944 1 555 0.000 y = 2.8 + 1.3
x
Quadratic 0.419 200.015 2 554 0.000 y = 15.8 14.5x + 2.5x 2
Cubic 0.556 230.901 3 553 0.000 y = 19.1 29.6x + 12.4x 2 1.4x 3
Compound 0.565 720.548 1 555 0.000 y = 11.5 0.44x
Power 0.929 7301.959 1 555 0.000 y = 3.2x 0.72
S 0.653 1042.726 1 555 0.000 ln y = 1.4 + 0.11
x
Growth 0.565 720.548 1 555 0.000 ln y = 2.4 0.81x
Exponential 0.565 720.548 1 555 0.000 ln y = ln 11.5 0.81x
Logistic 0.565 720.548 1 555 0.000
c1, as discussed in Chap.6. Moreover, the simple SEC model also requires less
input for predicting energy consumption, which is a major advantage for deriving
energy saving strategies. Therefore, this research prefers the model derived from
curve-estimation; with the SEC models and the values of the coefficients for the
tested CNC milling and machining centers listed in Table4.11. The SEC models
were then processed for validation.
64 4 Energy Efficiency Characterization of Manufacturing Processes
Residuals (kJ/cm3)
Fig.4.12Residuals plot of the inverse model for the Mori Seiki Dura Vertical 5100
1 0.947 y = 2.8 + 1.3/MRR
1 0.971 y = 2.8 + 1.3/MRR
1 0.997 y = 2.4 + 5.9/MRR
4.2.5Model Validation
The validation experiments were conducted without coolant, thus the above
SEC models for dry machining can be tested. Since the MRR is the only varia-
ble included in the SEC models, a different workpiece material (bright mild steel
1045) was introduced for testing at this stage. Moreover, the number of teeth
was initially tested as insignificant comparing with the impact of cutter diameter
(Sect. 4.2.2). Hence, a new cutter with 3 teeth, and diameter 32mm was used
to further validate the model. Other cutting parameters, such as cutting speed,
feed per rev per tooth and depth of cut, were changed to the ones which had not
been used at stage II. The validation results for the DMG DMU 60P are listed in
Table4.12. The results for other milling processes are presented in Appendix 4.
As the results showed, the difference between the predicted and measured
energy consumption remained under 5 % for the DMG DMU 60P. The valida-
tion results for other tested machine tools also showed less than 10 % discrepancy
between prediction and physical measurements. In other words, the SEC models
have achieved accuracy over 90 % for energy consumption prediction. Therefore,
the SEC models for tested milling processes can be concluded as accurate and
reliable.
Although there is a trend towards minimum coolant or lubricant use, wet cut is
still widely used in industry for milling processes. From an energy point of view,
applying coolant requires additional power to activate the coolant pump which is
generally at a constant level. Different machines provide different pressure of out-
put coolant. For some complex models, such as the DMG DMU 60P, an integrated
coolant filter system consumes extra energy once the coolant pump is enabled. All
the experiments have been repeated on 3 milling machine tools by using a low-
mineral, water-miscible coolant (Eco Cool 68 CF2, FUCHS Lubricants).
Table4.13 compares the unit energy consumption models between dry and wet
machining. The power profile presented in the table results from the measurement
of air cuts (running the cutting program without contacting the workpiece); the
blue and red curves are dry and wet air cut respectively. The gap between those
two curves is the power demand for applying coolant, which is approximately
equal to the difference between the coefficient c1 of SEC models for dry and wet
machining. For example, the coolant pump of the Mori Seiki Dura Vertical 5100
consumes 0.74kW, whereas the c1 value increases 0.675 from 1.344 to 2.019.
Table4.13Comparison of SEC models between dry and wet cutting environment (Kara and Li
2011)
Machine Model Power profile
Fadal VMC 4020 Dry SEC = 2.845 + 1.330
MRR =0.07kW
Weta SEC = 3.078 + 1.397
MRR
1400
1200
=0.74kW
1500
6700
=0.65kW
6400
aThe ANOVA results and model plots for wet machining are presented in Appendix 3
4.2 Energy Efficiency Characterization for Milling Processes 67
4.3.1Background
Studer S120
Year: 2007
Internal Grinding Wheel Spindle
Motor Power: 10 kW
Max RPM: 60000
Studer S40
Year: 2007
Internal Grinding Wheel Spindle
Motor Power: 7.5 kW
Max RPM: 60000
cutting mechanism, the formed chip or swarf is washed away by the circulated
coolant. Moreover, the coolant is emitted to the air which contains hazardous and
flammable substances. Hence, a coolant filter and an exhaust air filter are normally
connected to grinding machines. The three tested machine tools are displayed in
Fig.4.13. They cover a variety of machine age, spindle motor size, control system,
etc. All of the three machine tools were located in the Institute of Machine Tools and
Production Technology (IWF), Technische Universitt Braunschweig.
4.3.2Design of Experiments
carbon alloy steel 100Cr6 (62 HRC, Rockwell Hardness C scale), and a non-hard-
ened low carbon alloy steel 16MnCr5 (35 HRC).
) is the
In internal cylindrical grinding, the specific material removal rate (Qw
primary parameter which refers to the removal rate per unit width of grinding
(mm3/mms). The use of specific material removal rate reduces the number of vari-
ables and allows direct comparison of removal efficiency. The calculation of MRR
can be simplified as Eq.4.8.
MRR = Qw w (4.8)
During the experiments, the width of grinding (w) was equal to the thickness of
the workpiece (10mm), which is less than the thickness of the grinding wheel. It
was kept constant due to the use of a special workpiece holder. The turning and
milling experiments have proved that MRR has a decisive impact on the SEC.
Correspondingly, the design of experiments should seek as many levels as possible
for Qw in order to authentically characterize the trends between MRR and SEC.
Other cutting parameters are also believed to have a certain influence on the
energy consumption. For instance, the cutting speed (V) determines the rotation
speed of the grinding wheel (Ns), while it is also correlated with the type of grind-
ing wheel. Thus, the cutting speed was coupled with the type of grinding wheel.
But for a given grinding wheel, the cutting speed is held constant according to the
recommendations of the wheel suppliers. The rotation speed of the workpiece has
a ratio (qs) to the speed of grinding wheel. The ratio qs can vary from 30 to 90.
As the workpiece rotation speed is very slow comparing to the rotation speed of
grinding wheel, qs was initially considered as a screening factor.
Specific material removal volume (Vw ) is a unique process parameter for the
grinding case. It refers to the material removal volume per unit width of grind-
ing between two dressing processes (mm3/mm). The topography of grinding wheel
is constantly changing during the process due to the loss and self-sharpening of
grains, resulting in different cutting actions and frictions. Correspondingly, the
energy consumption may vary significantly during the grinding process, so Vw was
preceded for factor-screening.
The functions of cutting fluids are to reduce the temperature at the contacting
zone, to improve the surface finish, to increase the dimensional accuracy, and to avoid
excessive tool wear. There exists a variety of coolant used in industry i.e. mineral oil,
water-based emulsions, etc. Not only the material composition of coolant, but the
delivering of coolant can result in different lubricating and cooling effects, i.e. the
nozzle position, the flow rate of the coolant. However, from the energy consumption
perspective, the performance of coolant is out of the interest in this stage, except its
impacts on power requirement during the grinding process. Thus, both mineral oil
coolant and water-based emulsion are tested in the factor-screening step.
In order to estimate the impacts of specific material removal volume, one
grinding process was separated into a few runs. With the CBN grinding wheel,
400mm3/mm materials were removed in total before dressing. Therefore, the
experimenter configured each run by removing 100mm3/mm materials; and, after
repeating 4times, a dressing process was conducted. Figure4.14 plots the aver-
aged effective power against Vw .
4.3 Energy Efficiency Characterization for Grinding Processes 71
3800
3600
0 100 200 300 400 500
Specific Material Removal Volume (mm 3/mm)
4100
Reinecker qs = 30
Averaged Effective Power (W)
CBN 100Cr6 qs = 90
4000 =0.5
3900
3800
3700
100 200 300 400
Specific Material Removal Volume (mm3/mm)
Figure 4.14 has shown a clear power decreasing trend when increasing the
s pecific material removal volume. It is mainly because of the abrasive loss during
the grinding process, which reduces the number of effective cutting points as well
as the grinding forces. Therefore, Vw was considered as a design factor.
Two different speed ratios of workpiece and grinding wheel were used to test
its impact on the energy consumption, as shown in Fig.4.15. The results suggested
that the impacts of qs are less significant than Vw . Hence, it was held constant at
further steps.
A similar procedure was conducted by using mineral oil and water-based
lubrication. The results suggest that the type of coolant has a less significant
impact on energy consumption comparing to the specific material removal vol-
ume, as shown in Fig.4.16. Thus, the coolant type was considered as held-con-
stant factor at the following stages.
72 4 Energy Efficiency Characterization of Manufacturing Processes
5800
Studer S120 Mineral Oil
CBN 100Cr6
Averaged Effective Power (W)
=0.5 Emulsion
5600 qs = 60
5400
5200
0 100 200 300 400 500
Specific Material Removal Volume (mm3/mm)
Design Design
Factors Factors
Cutting speed
Workpiece
material Specific material
removal volume
Specific
Energy
Tool wear Width of grinding Consumption
After factor-screening, all the involved factors were classified into different
groups as displayed in the Ishikawa diagram in Fig.4.17.
As mentioned before, there are many constraints for defining the levels of pro-
cess variables. Instead of using RSM for design of experiments, the full factorial
experiments were used in the case of grinding. For most of the factors, the range
of variance is very limited. From the test of turning and milling process, the mate-
rial removal rate has been proved as a key factor (Sects.4.1 and 4.2). Previous
4.3 Energy Efficiency Characterization for Grinding Processes 73
researches about grinding forces also showed the decisive impacts of Qw on specific
grinding energy (Malkin 1975). Thus, the levels of variable Qw were maximized.
Table4.14 lists the levels of each design factor for the tested grinding machine tools.
4.3.3Experiment Details
The experiment for grinding process was kept simple and straightforward. The
power curve of an exemplary run was shown in Fig.4.18.
The first power peak reflects the acceleration of the grinding wheel spindle. At
the same time, the grinding wheel rapidly travelled towards the workpiece until
reaching the safety distance, then slowed down to the same speed as actual grind-
ing. In practice, the safety distance is kept minimal, because it refers to a non-
value-adding period. Nevertheless, the safety distance was kept relatively long in
this research to help the data analysis by separating the energy consumed during
the grinding period. The instantaneous power started to increase once the grind-
ing wheel contacted with the workpiece surface. Then the power consumption
decreases along the grinding process due to the topography change of the grind-
ing wheel. In each run, 100mm3/mm or 50mm3/mm of workpiece materials were
removed by the CBN wheel or Al2O3 respectively. After repeating the procedure
for three or four times, the grinding wheel was dressed with a fixed dressing
parameter. During the experiments, a mineral oil based emulsion was employed
as coolant. The emulsion had a concentration of 8.0 % and kinematic viscosity of
1.1m2/s at 40c. The coolant was delivered as a flood through a tangential nozzle
into the contact zone.
4.3.4Regression Results
First, the power demand during the grinding period was separated from the raw data
file, as shown in Fig.4.18a. The SEC then can be calculated according to Eq.4.9.
74 4 Energy Efficiency Characterization of Manufacturing Processes
6000
4000
2000
0 Time
1 201 401 601 801 1001 1201 1401 (0.1s)
(b)
Nwheel
Nworkpiece
10 mm
Fig.4.18Exemplary power curve of one grinding experiment and illustrations of internal grind-
ing operations. a Exemplary power curve of one grinding experiment. b Schematic illustration of
internal grinding operations
E Pi d t
SEC = = (4.9)
Q Vw w
Then, each design factor was processed with curve-estimation in relation to SEC.
The material removal rate (MRR) was used instead of specific material removal
rate, since MRR is a more comparable process parameter. Other indirect process
parameters have also been tested, such as the rotation speed of grinding wheel
spindle (Ns, rpm), the feed speed of grinding wheel (vfr, mm/min), and the feed
length of grinding wheel (lfr, mm). Similar to turning and milling, the MRR has
showed the highest R-square value comparing to the other factors.
However, different model shapes have achieved R-square value over 0.9. This is
mainly due to the limited range of material removal rate. Unlike turning or milling
process, the observed data for grinding can only plotted in a segment of the curve
4.3 Energy Efficiency Characterization for Grinding Processes 75
SEC (kJ/cm3)
MRR
(cm3/s)
Fig.4.19Model plot between MRR and SEC for the Studer S120
in which a number of models can fit well. The high steepness of the curve can
even be approximated as a linear line.
Regardless, the inverse model still has resulted in the highest R-square value
and the largest F-value among all the 11 available models. The statistical results
have suggested that the inverse model is the best-fit model after curve-estimation.
The model plot and statistical results for the Studer S120 are presented in Fig.4.19
and Table4.15 respectively. The ANOVA analysis of inverse model and model
plots for the other tested machine tools can be found in Appendix 3.
The residual analyses of the inverse models show mixed results. The residuals
plot for the Reinecker ISA 110 shows an unclear trend (Fig.4.20), whereas there
is an arrow shape in the one for the Studer S40 (Fig.4.21). Thus, stepwise multi-
ple-linear regression was performed afterwards.
All the collected data was processed for a stepwise multiple-linear regression.
As listed in Table4.16, the model for the Reinecker only accepts MRR as a signif-
icant factor, whereas the models for other two machines disagree with each other.
The R-square values were only increased by less than 2 % after introducing new
factors. Therefore, the other process parameters can be concluded negligible for
estimating SEC.
Based on above statistical analysis, the SEC model for tested internal cylindri-
cal grinding processes can be summarized in Table4.17. The grinding SEC mod-
els join the same shape with the ones for turning and milling processes, as shown
in Eq.4.4.
76 4 Energy Efficiency Characterization of Manufacturing Processes
Table4.15Model summary of curve-estimation between MRR and SEC for the Studer S120
Model type Model summary Equation
R2 F-value DF1 DF2 P-value
Linear 0.718 165.575 1 65 0.000 y = 1648 95509x
Logarithmic 0.921 754.243 1 65 0.000 y = 3220 826 ln x
Inverse 0.996 17,335.020 1 65 0.000 y = 70 + 4.8/x
Quadratic 0.934 452.098 2 64 0.000 y = 2534 332868x
+ 1.2 107 x 2
Cubic 0.986 1508.654 3 63 0.000 y = 3279 6.6 105
x + 5.3 107 x 2
1.4 109 x 3
Compound 0.916 705.686 1 65 0.000 y = 1939 6.70.051x
Power 0.995 14,352.768 1 65 0.000 y = 7.9 x 0.92
S 0.913 681.122 1 65 0.000 lny = 5.8 + 0.005/x
Growth 0.916 705.686 1 65 0.000 ln y = 7.6 115.5x
Exponential 0.916 705.686 1 65 0.000 ln y = ln 1939 115.5x
Logistic 0.916 705.686 1 65 0.000
Residuals (kJ/cm3)
Fig.4.20Residuals plot of the inverse model for the Reinecker ISA 110
4.3.5Model Validation
Residuals (kJ/cm3)
1 0.988 y = 135 + 6.5/MRR
NS refers to spindle rotation speed (rev/min); vfr refers to feed speed (mm/min); HRC refers to
Rockwell hardness; lfr refers to feed length (mm)
complete results are presented in Appendix 4. As the results show, the errors of the
predictions remained less than 10 % of the total energy consumption. Therefore,
the derived SEC models have been validated with a high accuracy.
4.4.1Background
Heating unit
Clamping
Injection
cylinder
cylinder
Hydraulic
Process controller
4.4.2Design of Experiments
Unlike the metal machining processes, there are several stages involved in one
injection molding cycle; they include plasticization, filling, packing, cooling, and
part ejection (Rees 2002). First, the plastic granules or pellets are melted in a heated
chamber in which several heaters are switched on and off to maintain the desired
temperature; then, the molten plastic is injected into the cavity, while material
shrinks naturally; thus, the additional space in the cavity is filled during packing
stage; once the cavity is filled completely, the molten plastics solidify with the help
of a cooling system; finally, the part is ejected from the mold. There are numerous
factors affecting the process, which can be grouped into the following categories:
Type of material: generally referring to melting temperatures and density;
Mold design: i.e. shape, size, number of cavities, wall thickness, etc.;
Plasticizing unit: number of heaters, temperatures, decompression pressure and
speed, etc.;
80 4 Energy Efficiency Characterization of Manufacturing Processes
BOY 15S
Year: 1978
Max Pressure: 140 bar
BOY 15
Year: 1976
BattenfeldBA500CD
Year: 1990s
Max Pressure: 1484 bar
Clamping force: 500 kN
Injection and holding unit: i.e. injection pressure and speed, holding pressure, etc.;
Clamping unit: i.e. clamping pressure, mold close and open speed;
Cooling unit: cooling water flow rate;
Ejecting unit: ejection speed and pressure;
Timers: injection and holding time, cooling time, and dwell time;
Operation environment: i.e. temperature, humidity, etc.
Correspondingly, there are a large number of control options to configure the
above mentioned process parameters. Figure4.24 shows an example of BOY 15S.
The switches for pressures and speeds set a certain percentage of the maximum
value. The timer can numerically configure the injection and holding time, the
cooling time, and the dwell time.
Different materials feature different melting temperature and heat capacity.
From the energy point of view, the energy consumption of heating units would
correspond to the use of different materials. Normally, the heaters are set at the
lowest possible temperature to melt the plastics while achieving adequate vis-
cosity. Moreover, different heat capacity results in different time for melting and
cooling the material. Thus, the material type was initially determined as a design
factor, and the temperature was coupled with it.
4.4 Energy Efficiency Characterization for Injection Molding Processes 81
(a) (b)
Part A Part B
Insert A Insert B 30 mm 36 mm
w 9mm w 13mm
Fig.4.25Photograph of the inserts and injected parts on the BOY 15S. a Photograph of inserts.
b Photograph of parts (Li etal. 2015)
The mold design also has a significant impact on this process (Rees 2002). The
volume, wall thickness (w), cavity numbers were suggested to limit the selection
of pressures, speeds, as well as the cycle time. However, mold making is an expen-
sive and time-consuming process due to the high requirements of surface finish
and dimensional accuracy. CNC machining is currently the main means of mak-
ing the molding tools, owing to the accuracy and flexibility of the machine tools.
For the molds with complex shape, electrical discharge machining (EDM) is also
used. All the tested machines featured a two-cavity, twoplate mold. Two inserts
were designed for this research, as shown in Fig.4.25. They featured a basic round
shape with different diameters and thickness (w). This design allowed the experi-
menters to change the volume of injected material. When both two inserts were
82 4 Energy Efficiency Characterization of Manufacturing Processes
installed normally, the maximum injection volume was achieved. The cavity can
also be blocked in order to change the injection volume.
The cycle time is another important indicator which is used to assess the pro-
ductivity and economic performance of the process. It generally consists of three
elements: injection and holding time, cooling time, and dwell time. Ideally, the
total cycle time should be reduced to a minimal level. Thus, the dwell time, which
is the pausing period between two cycles, was kept constant as 2s throughout the
experiments. The other two elements determine the total cycle time. Therefore, the
injection and holding time and cooling time were considered as design factors.
There are a number of parameters for different speed and pressure. Obviously,
it is inefficient to change each one of them. Some of the parameters are kept con-
stant in practice, such as ejection speed and pressure, mold open and close speed,
etc. Those parameters are set at the maximum value in order to minimize the cycle
time. However, it was difficult to screen all the other parameters relating to the
speed and pressure (Qureshi etal. 2012). This was due to the power consumptions
of the three heaters, which switch on and off at a different pace. As a result, the
accuracy of the analysis was influenced significantly. Alternatively, the speed and
pressure parameters were combined as one factor. It was reasonable to assume that
the higher speed and pressure may require more energy; and the combined factor
would result even more significant trend. For example, if higher injection speed or
higher clamping pressure require more power individually, setting both parameters
at the high level would consumes much more power. Therefore, the involved fac-
tors can be summarized in an Ishikawa diagram, see in Fig.4.26.
Design Design
Factors Factors
Pressure and Speed
( Clamping pressure Material
Injection pressure
Injection volume
Decompression pressure
Holding pressure Injection and holding
Injection speed time
Decompression speed
Cooling time
Dosing speed) Specific
Energy
Cooling water flow rate Dwell Time Consumption
Nuisance Held-constant
Factors Factors
As mentioned before, the plastics have firstly been melted and then solidified.
The shrinkage during the cooling stage is normal and expected. This drawback
has attracted intensive research to predict and manage this phenomenon (Fischer
2003). Since this research needs observed data for input, it is impractical to meas-
ure the volume change. Alternatively, the mass of the processed material (m) is
constant throughout plasticization and solidification. The injected part can be
easily weighted as shown in Sect.3.1.4. Therefore, the model response SEC for
injection molding process is the total energy consumption of producing 1g of part
(kJ/g); see the equation in Eq.4.10.
E Pi d t
SEC = = (4.10)
m m
Throughput rate is another important indicator for injection molding process,
which suggests the process productivity. It can be considered as similar to MRR
for material removal processes. The calculation of throughput rate (, g/s) is
straight forward as Eq.4.11.
m m
m = = (4.11)
T ti + t c + t d
where ti refers to the injection and holding time; tc refers to the cooling time; and,
td refers to the dwell time between two cycles.
After a few trial runs, the levels of each design factor were determined accord-
ingly for the three tested materials, as shown Table4.19. The possible wide range
has been obtained in which the injected parts were neither incomplete nor with
excessive flushes. Similar to the other processes, the experiment schedules were
generated through the use of MiniTab.
4.4.3Experiment Details
From the empirical modelling point of view, each combination of process vari-
ables should be repeated several times. Hence, the machines were running at
automatic mode, and produced at least 10 parts with each process configuration.
Figure4.27 shows an exemplary power curve over 11 cycles. Each cycle has then
been separated and processed to calculate the SEC according to Eq.4.10.
4.4.4Regression Results
The similar statistical analysis was conducted here. Each process parameter was
processed with curve-estimation. Similar to the results of material removal pro-
cesses, the throughput with an inverse model has resulted in the highest R-square
value and largest F-value. Figure4.28 and Table4.20 presents the model plot
84 4 Energy Efficiency Characterization of Manufacturing Processes
and the model summary for the BOY 15S respectively. The ANOVA analysis of
inverse model and model plots for the other tested machine tools can be found in
Appendix 3.
The residual analysis also suggest that the inverse model is a good one, since
there is no clear trend observed in the residuals plot, see Fig.4.29.
Based on above statistical analysis, the SEC model for injection molding pro-
cess can be written in a form as Eq.4.12. The values of coefficient c0 and c1 for the
tested injection moulding processes are summarized in Table4.21.
c1
SEC = c0 + (4.12)
m
4.4.5Model Validation
Despite the strong statistical evidence, additional runs were conducted on BOY
15S to further validate the derived model. Since the models require throughput
rate as input, the mass of injected parts should be estimated. This can be done by
multiplying the volume of cavities with material density. The cycle time can be
4.4 Energy Efficiency Characterization for Injection Molding Processes 85
9000
6000
3000
0 Time
0.1s
Mould Open
12000
Mould Closure
Mould Open
10000
Effective power (W)
8000
Decompression
6000
Heater
4000
2000
Time
0
0.1s
Injection and Cooling time
holding time
Fig.4.27Exemplary power curve of the tested injection molding process (Li etal. 2015)
easily estimated based on the configuration of injection and holding time, cool-
ing time and dwell time. It should be noted that the estimation of throughput con-
tains errors, which assumes that there is no shrinkage or flush of the injected part,
and the machine timing is accurate. For the validation runs, the insert A (see in
Fig.4.25) was turned over in order to block one cavity, so the volume of injected
part was different than the ones for model development. The results are listed in
Table4.22.
86 4 Energy Efficiency Characterization of Manufacturing Processes
SEC (kJ/g)
.
m (g/s)
Fig.4.28Model plot between throughput and SEC for the BOY 15S
Table4.20Model summary of curve-estimation between throughput and SEC for the BOY 15S
Model type Model summary Equation
R2 F-value DF1 DF2 P-value
Linear 0.72h 2585.25 1 998 0.000 y = 21.7 30.1x
Logarithmic 0.917 11,004.06 1 998 0.000 y = 0.18 9.3 ln x
Inverse 0.982 54,054.51 1 998 0.000 y = 3.6 + 1.9/x
Quadratic 0.917 5529.60 2 997 0.000 y = 30 94x + 91x 2
Cubic 0.970 10,714.44 3 996 0.000 y = 38 186x + 383x 2 263x 3
Compound 0.860 6136.48 1 998 0.000 y = 24 0.09x
Power 0.971 33,647.59 1 998 0.000 y = 4.4 x 0.703
S 0.928 12,887.542 1 998 0.000 ln y = 1.8 + 0.138/x
Growth 0.860 6138.48 1 998 0.000 ln y = 3.2 2.4x
Exponential 0.860 6138.48 1 998 0.000 ln y = ln 24 2.4x
Logistic 0.860 6138.48 1 998 0.000
Residuals (kJ/g)
.
m (g/s)
4.5.1Background
Fig.4.30Photograph of
tested ram EDM machines.
a SODICK A3R. b AGIE
TRON 30L
4.5.2Design of Experiments
tested. Since each design factor has different levels of variances, a full factorial
experiment design was used in this case. Each combination of process parameters
were repeated for 5times. Table4.23 lists the tested levels of variances.
4.5.3Experiment Details
Fig.4.32Photograph of one
EDM experiment
92 4 Energy Efficiency Characterization of Manufacturing Processes
Machining Machining
start end
Cooling system
switch-on Cooling system Cooling system
switch-on Cooling
switch-off
system
Power (w)
switch-off
Time (0.1s)
control the power and duration of each pulse. The total machining energy con-
sumption is calculated as Eq.4.13.
Etotal = PAve.Mach (t1 t0 ) (4.13)
where PAve.Mach is the average machining power; t0 is the machining start time;
and, t1 is the machining end time.
As mentioned before, each run of experiment has been kept constant for 5min,
and the final depth-of-cut has also been precisely recorded. Then, the total material
removal volume, Qvolume can be calculated depending on the shape of electrode
and the final depth-of-cut. Hence, the SEC and MRR of each experiment run can
be derived according to Eqs.4.14 and 4.15.
Etotal
SEC = PAve.Mach (4.14)
Qvolume
Qvolume
MRR = (4.15)
t1 t 0
Notably, the cooling system and the dielectric fluid pump remained active after
the end of machining process. It took approximately 1min until the power was
back to the fixed level. The amount of energy consumption due to this reason was
almost constant disregarding the duration of machining processes or the mate-
rial removal volume. In other words, it is independent from the design factors.
Therefore, this share of energy consumption was purposely excluded for the mod-
elling stage in Sect.4.5.4. Alternatively, it should be added as a fixed amount for
each machining process.
4.5 Energy Efficiency Characterization for Electrical Discharge Machining 93
4.5.4Regression Results
At this stage, all the observed SEC and MRR values were processed with SPSS
for modelling and regression analysis according to the proposed methodology in
Chap.3.
Firstly, the relationship between SEC and process parameters was analyzed
by using the curve estimation. This analysis investigates the main effect of each
process parameter or a combined process parameter, and it determines the best-
fit model among the 11 available model types. All the design factors showed a
poor ability to individually describe the variance of SEC. The R2-value for these
models is less than 0.3. Like conventional material removal processes, the MRR,
a combined process parameter, has a decisive influence on the SEC of EDM
processes. As shown in Fig.4.34, the inverse model is the best-fit between the
observed SEC and MRR on SODICK A3R, whose R2-value is 0.99. The same
trend has been observed for the AGIE TRON 30L. The derived SEC models are
summarized in Table4.24.
SEC (kJ/cm 3)
Fig.4.34The curve estimation between SEC and MRR of SODICK A3R (Li and Kara 2015)
94 4 Energy Efficiency Characterization of Manufacturing Processes
Since the MRR has been proved as a decisive factor for SEC models, the pre-
diction of MRR becomes equally important in order to predict energy consump-
tion of EDM processes. Owing to the stochastic nature of EDM processes, it is
challenging to derive models for predicting the MRR. While previous researches
were mainly focused on maximizing MRR, a number of modelling efforts have
been mentioned by Hinduja and Kunieda (2013). For example, Yahya and
Manning (2004) derived a mathematical MRR model by using dimensional analy-
sis; Wang and Tsai (2001) provide a semi-empirical model to predict the MRR of
work and tool; Torres etal. (2015) provided an empirical MRR model of machin-
ing hard-to-machine alloys. However, the model form appears different among
those approaches, and the derived empirical coefficients are limited to the tested
combination of electrode and workpiece material. Therefore, the observed MRR
through the experiments were used again to further characterize the relationship
between MRR and process parameters.
The process parameters (e.g. peak discharge current, on-time, etc.) were first
used to describe the observed MRR. The curve estimation suggests that there is
a significant correlation between the process parameters and the MRR. However,
the R-square value is low which means the MRR is affected by a number of pro-
cess parameters. Alternatively, two theoretical MRR models from previous ana-
lytical studies were selected to test the model validity, and to derive coefficients
for prediction purpose. The theoretical model 1 is based on Yahya and Mannings
dimensional analysis (2004), whereas the theoretical model 2 was based on the
semi-empirical study by Wang and Tsai (2001). The observed MRR and process
parameters were then processed with ANOVA analysis by SPSS, as summarized
in Table4.25.
As the R-square value suggests, both MRR models are capable of describing
the observed variance. The theoretical model 2 has a higher R-square value, which
is mainly due to the consideration of a higher order of process variables. However,
the coefficients of the theoretical model 2 appear less logical than the ones of the
theoretical model 1. For example, the aluminum workpiece, comparing to mild
steel, is easier to machine and has a lower melting temperature. The experimental
results confirmed that the MRR of machining aluminum is higher than mild steel.
The coefficient A1 in the theoretical model 1 agrees with the trend, whereas the
one in the theoretical model 2 shows an opposite trend. In addition, the R-square
difference between model 1 and 2 is marginal. Therefore, the theoretical model 1
is preferred which is also more practical.
4.5 Energy Efficiency Characterization for Electrical Discharge Machining 95
Table4.25Derived MRR models for the tested combinations of electrode and workpiece
materials
Theoretical model 1 MRR = A1 U I ton F
Electrode Copper Graphite
Workpiece Steel Aluminum Steel Aluminum
A1 1.660108 4.993108 2.429108 5.707108
R-square 0.936 0.964 0.897 0.976
Theoretical model 2 2 3
MRR = A1 A2 + A3 tratio + A4 tratio + A5 tratio U I ton F
Electrode Copper Graphite
Workpiece Steel Aluminum Steel Aluminum
A1 1.415109 3.5391010 0.078 0.003
A2 9.371 301.525 0.026 1.445
A3 0.815 180.250 0.080 4.478
A4 0.690 50.911 0.077 4.334
A5 0.065 3.625 0.023 1.300
R-square 0.988 0.979 0.963 0.982
Ai refers to the model coefficient; U refers to the no-load voltage (in the unit of V); I refers to the
peak discharge (in the unit of A); ton refers to the on time (in the unit of s); F refers to the fre-
quency; tratio refers to the ratio between on time and off time; the unit of MRR is mm3/min
4.5.5Model Validation
Besides the statistical proofs presented in Sect.4.5.4, additional runs were con-
ducted on both tested machines to further validate the accuracy of derived empiri-
cal models. Comparing to the process parameters used in the stage II, two different
settings were used with five repetitions. The MRR was estimated according the
theoretical model 1 (see in Table4.25). The predicted SEC was compared with
the measured average SEC as listed in Table4.26. The error margin of the energy
consumption prediction is calculated, which is less than 10% for all the validation
tests. Therefore, the derived SEC and MRR models are accurate enough for pre-
dicting the energy consumption of EDM processes.
4.6Summary
This chapter has validated the proposed methodologies to characterize the energy
efficiency on variant manufacturing processes including turning, milling, grind-
ing, injection molding and electrical discharge machining. Among all the tested
cases, a reverse trend has been consistently observed between the specific energy
96 4 Energy Efficiency Characterization of Manufacturing Processes
consumption and process rate (e.g. material removal rate and throughput). The
derived SEC models offer a great ability to predict energy consumption of a given
process with an accuracy of over 90 %. The observed data as listed in Table4.27
can be accessed via the link (http://extras.springer.com).
References 97
References
Abbas NM, Solomon DG, Bahari F (2007) A review on current research trends in eletrical dis-
charge machining. Int J Mach Tools Manuf 47:12141228
Armarego EJA, Wiriyacosol S (1978) Oblique machining with triangular form toolsI.
Theoretical investigation. Int J Mach Tool Des Res 18(2):6780
Fischer JM (2003) Handbook of molded part shrinkage and warpage, 1st edn. William Andrew,
Norwich
Gutowski T, Dahmus J, Thiriez A (2006) Electrical energy requirements for manufacturing pro-
cesses. In: 13th CIRP international conference on life cycle engineering. Leuven
Hinduja S, Kunieda M (2013) Modelling of ECM and EDM processes. CIRP Ann Manuf
Technol 62(2):775797
Ho KH, Newman ST (2003) State of the art electrical discharge machining (EDM). Int J Mach
Tools Manuf 43:12871300
Kalpakjian S, Schmid SR (2005) Manufacturing engineering and technology, 5th edn. Prentice
Hall, Englewood Cliffs
Kara S, Li W (2011) Unit process energy consumption models for material removal processes.
CIRP AnnManuf Technol 60(1):3740
Li W, Kara S (2011) An empirical model for predicting energy consumption of manufactur-
ing processes: a case of turning process. In: Proceedings of the Institution of Mechanical
Engineers, Part B. J Eng Manuf 225(9):16361646
Li W, Kara S (2015) Characterising energy efficiency of electrical discharge machining (EDM)
processes. Procedia CIRP 29:263268
Li W, Kara S, Qureshi F (2015) Characterising energy and eco-efficiency of injection moulding
processes. Int J Sustain Eng 8(1):5565
Malkin S (1975) Specific energy and mechanisms in abrasive processes. In: Proceedings of third
North American metalworking research conference. Carnegie Press, Lancaster
Malkin S, Guo C (2008) Grinding technologtheory and applications of machining with abra-
sives, 2nd edn. Industrial Press, New York
Oxley PLB (1998) Development and application of a predictive machining theory. J Mach Sci
Technol 2(2):165189
Qureshi F, Li W, Kara S, Herrmann C (2012) Unit process energy consumption models for mate-
rial addition processes: a case of the injection molding process. In: 19th CIRP international
conference on life cycle engineering. Berkeley
Rees H (2002) Mold engineering, 2nd edn. Hanser Gardner Publications, Cincinnati
Rosato DV (2000) Injection molding handbook, 3rd edn. Springer, Berlin
SECO Tools (2009) MN2009-turning: the machining navigator catalogue
Sreejith PS, Ngoi BKA (2000) Dry machining: machining of the future. J Mater Process Technol
101(1):287291
Torres A, Puertas I, Luis CJ (2015) Modelling of surface finish, electrode wear and mate-
rial removal rate in electrical discharge machining of hart-to-machine alloys. Precis Eng
40:3345
Wang PJ, Tsai KM (2001) Semi-empirical model on work removal and tool wear in electrical
discharge machining. J Mater Process Technol 114:117
Yahya A, Manning CD (2004) Determination of material removal rate of an electro-discharge
machine using dimensional analysis. J Phys D Appl Phys 37:14671471
Chapter 5
Eco-efficiency of Manufacturing Processes
Fig.5.1Environmental 0.012
impact (CO2 Fossil) against
MRR (Colchester Tornado Australia
A50)
0.009 USA
European Union
0.006
CO2 Fossil
Critical Region
0.003
MRR
0 3
0 0.5 1 1.5 2 2.5 3 (cm /s)
difference. Of course, other quality measures are important for injection moldings,
such as surface finish, shape complexity, and dimensional accuracy. However,
these measures are product specific, and it is difficult to be objective when com-
paring among different products. Moreover, these measures are mainly relying on
the mold design and production. From a unit process point of view, theses quality
measures are not comparable among different injected parts. Thus, these measures
cannot be compared equally. Alternatively, it would be helpful to develop a matrix
for evaluating the value of the process including above mentioned requirements,
meanwhile each measure can be weighted according to customers preferences.
To develop such a matrix requires comprehensive surveys about the capacity of
the injection molding processes, the general customer requirements, the variety
of injected parts, etc. At this stage, the weight is used for representing the value
of the injection molding processes. For a given shape of injected part, the value of
injection molding process can be increased by maximizing the parts per injection
cycle. In that sense, the total weight of injected parts per cycle is maximized, so is
the value of the process.
CO2 Fossil
0.006
Critical Region
0.003
Throughput
0 (g/s)
0 0.2 0.4 0.6 0.8 1
Unlike turning and milling processes, the surface integrity has the highest priority
for grinding processes. Thus, the surface roughness should be included to value
this process, which is also extremely sensitive to the topography of the grind-
ing wheel or any changes of process parameters (Jawahir etal. 2011). The sur-
face roughness Rz was monitored during the additional grinding experiments with
HOMMEL-ETAMIC T1000 basic, as shown in Fig.5.3.
The additional experiments were conducted on the Studer S120 under differ-
ent material removal rate and different material removal volume. The experiment
104 5 Eco-efficiency of Manufacturing Processes
procedure was very similar to the experiments during model development, as pre-
sented in Sect.4.3. Both CBN and Al2O3 wheels were tested to grind 100Cr6. The
surface roughness Rz (m) was measured at four different points at the work piece.
The measured surface roughness was plotted over the energy consumption trend
in Figs.5.4 and 5.5 for CBN and Al2O3 respectively. As shown in both figures,
the surface roughness measurement after grinding the first 100mm3/mm remains
relatively constant due to the effective dressing process (the lower dash line).
However, there is a general trend that the surface roughness worsens as MRR
increases, especially when grinding with Al2O3 grinding wheels. It clearly shows a
trade-off between energy consumptions and surface roughness.
For a given MRR, increasing specific material removal volume (Vw) also
degrades the surface roughness. As shown in the Figures, the longer color bar
indicates the requirement for more frequent dressing in order to maintain constant
surface finish. The relationship between surface roughness (RZ) and process param-
eters (Vw and Qw) were given empirically as shown in Eq.5.1 (R2=0.908).
RZ = 0 + 1 QW VW (5.1)
According to the trade-off analysis, there is a cross region resulted in the Figs.5.4
and 5.5. It suggests that when MRR is around 0.01cm3/s, grinding with CBN wheel
will result in low energy consumption, high surface finish and relatively low tool
wear rate. A similar point was found for Al2O3 grinding wheels as 0.0075cm3/s.
They can be recommended as the first-choice of material removal rate.
Besides the electrical energy consumption during the grinding process, it requires
dressing process to keep the sharpness of the grinding wheel. As the color bar
in Figs.5.4 and 5.5 shows, the grinding wheel lost the prefect sharpness during
5.3 Eco-efficiency of Grinding Processes 105
5
2000
Specificc Energy Consumptioon (kJ/cm3)
4
Surrface Roughness Rz (
1500
VW = 500 mm3/mm
3
1000
2
m)
500 1
VW = 100 mm3/mm
0
0 MRR
0 005
0.005 0 010
0.010 0 015
0.015 0 020
0.020 (cm3/s)
Fig.5.4Trade-off between energy consumption and surface roughness (CBN wheel on the
Studer S120)
the abrasive machining process. Once the surface roughness exceeds the desired
requirement, the dressing process should be performed, which consumes electric-
ity, coolant, grinding wheel and dresser. During the additional experiments, the
radial grinding wheel wear rs was also measured. After each infeed, the grinding
wheel was ground in a thin steel plate (0.25mm); then, the grinding wheel profile
was measured through the thin steel plates with a measuring probe, as shown in
Fig.5.6.
The empirical relationship between dressing amount (rs) and process param-
eters were derived and displayed in Eq.5.2 (R2=0.904).
rs = 0 + 1 QW VW (5.2)
The information can be used to determine the amount of grinding wheel to be
removed during dressing process. It can be considered as the tool consumption of
the unit process. Thus, tool consumption can be theoretically included for the eco-
efficiency evaluation.
106 5 Eco-efficiency of Manufacturing Processes
5
2000 VW = 200 mm3/mm
Specificc Energy Consumption (kJ/cm3)
4
1000
2
500 1
VW = 50 mm3/mm
0
0 MRR
0.005
0 005 0.010
0 010 0.015
0 015 0.020
0 020 (cm3/s)
Fig.5.5Trade-off between energy consumption and surface roughness (Al2O3 wheel on the
Studer S120)
A screening Life Cycle Assessment (LCA) was conducted to assess the envi-
ronmental impacts of grinding wheels, including 4 steps: goal and scope, inven-
tory analysis, impact assessment, and interpretation.
The functional unit of this analysis was defined as raw materials and disposal
of 1kg of CBN and Al2O3 grinding wheel. Owing to the limited information, the
analysis was only focusing on the raw material and end of life stage. Since the use
phase of grinding wheel is the grinding process and dressing process, the electric-
ity consumption has already been considered from the unit process point of view,
thus it was excluded for the grinding wheel analysis.
The CBN grinding wheel mainly consists of the cubic form of boron nitride,
or called as -BN. Other material and elements are added to change its color, con-
ductivity and toughness, such as Li3N, Be, Si, C, or P (Wentorf 1961; Greim and
Schwetz 2006). The -BN was made from boron acid with urea through a hot-
pressing process (46GPa, 14001700C). The chemical reaction is listed below:
5.3 Eco-efficiency of Grinding Processes 107
>1000 C
B2 O3 + CO(NH2 )2 2BN + CO2 + 2H2 O
According to the molecular weight, the amount of raw materials for producing
1kg -BN powder can be estimated as 1.4 and 1.2kg for boron acid and urea
respectively. The Al2O3 grinding wheel consists of 98% of Al2O3, 1.5% of TiO2,
and other bonding materials (Klocke and Knig 2005).
The production of grinding wheel generally requires 4 stages. The process
begins with selecting and mixing precise quantities of abrasives, bond materials,
and additives according to a specific formula. Then, the mixture is poured into a
mold and pressed with a pressure in the range of 1005000psi. After molding,
the vitrified grinding wheels are fired to temperatures between 927 and 1260C
in order to melt the binder around the abrasives and to convert the wheels to a
form that resist the heat and solvents encountered during grinding. Finishing was
the last step to correct thickness or parallel wheel sides, or to create special con-
tours on the side or circumference of the wheel (Borkowski and Szymaski 1992;
Marinescu etal. 2000). However, the current information is insufficient to quantify
the energy requirements during the production stage. In addition, the use of prob-
lematic chemicals and the related emissions remains unidentified. Therefore, fur-
ther investigation into the production of grinding wheels is recommended.
The removed abrasives and bonding materials are flooded away with grinding
swarf. It is difficult and uneconomical to recycle the removed grinding wheels.
Thus, the disposal of grinding wheel was assumed as 100% land fill.
108 5 Eco-efficiency of Manufacturing Processes
The inventory analysis produces the Life Cycle Inventory (LCI) results based
on the above literature and the Ecoinvent 2.2 database. It includes the inputs and
outputs of raw materials, emissions and wastes where results of the amount of car-
bon dioxide fossil emission have been extracted. A 7.94 and 1.81kg fossil fuel
CO2 emission was calculated for the CBN and Al2O3 grinding wheel respectively.
Each dressing process only removes few micro meters of grinding wheel. For
removing 1kg of raw material, the loss of grinding wheel is in the range from 3 to
10g (the density of CBN wheel is 3.48g/cm3). Comparing to the electricity con-
sumption of grinding and dressing process, the environmental impacts of grinding
wheel consumption appears to be relatively insignificant. This exclusion requires
further validation once a full scale LCA of grinding wheels becomes available.
Unlike turning and milling processes, coolant is essential for grinding process due
to the substantial amount of heat generated at the contacting zone. As discussed in
Sect.3.2, the environmental impacts of applying coolant are due to the electricity con-
sumption for flow generation, as well as due to the coolant loss through the activities.
Since the coolant pump requires constant power, the electricity consumption due to
coolant pump can be easily estimated. The coolant loss mainly occurs along with the
flood of grinding swarf, which is depending on the operation conditions. Assumptions
are necessary to estimate the coolant loss during the grinding process. Comparing to
the embodied energy of coolant other environmental impacts are also important, such
as the fluid splashing, improper disposal, human health problem due to the exposure
to the coolant, etc. As a consequent, there has been an occupational exposure limit
(OEL) of 10mg/m3 for the sum of aerosol and vapor concentration of coolants with
a flash point above 100C since 1996 (Schiefer 2000). Currently, air exhauster is
essential for grinding machine tools to minimize the human exposure to the hazard-
ous substances. However, the air exhauster cannot completely eliminate the hazardous
substances (Dettmer 2006). In order to evaluate the human health related issues, the
investigation of human exposure to coolant is recommended for future work.
In this research, the electricity due to the coolant pump was considered as the
main environmental impact of applying coolant. In addition, the coolant loss was
assumed to be between 5 and 30% of the fill quantity (Clarens etal. 2008). The air
exhauster was assumed to be functioning properly, thus human health related issues
can be excluded. According to (Winter etal. 2012), 0.168kg fossil fuel CO2 is emit-
ted to produce 1kg mineral oil based emulsion; and, another 0.730kg fossil fuel
CO2 is emitted during the end-of-life phase assuming a combustion in a hazardous
waste incinerator (no credit was given for the recovered energy). Thus a rough esti-
mation can be obtained for the CO2 emissions due to the consumed coolant.
The red lines in Fig.5.7 show an evenly decreasing trend of the CO2 Fossil
caused by an increasing MRR. This can be explained by the reduced process time;
indeed the required process energy is higher but simultaneously the share of base
load is reduced by the shorter process time. Chasing the high surface finish (low
RZ value) results in higher CO2 Fossil. After a certain point, the environmental
impact goes up exponentially. Therefore, a critical region has been defined as
shown in Fig.5.7.
40
32
O2 Fossil
O2 Fossil
24
CO
CO
16
0
0.005 0.007 0.009 0.011 0.013 0.015
( 3/s))
MRR(cm
40
32
Fossil
Fossil
24
4
3
CO2 F
CO2 F
(kg per grindin
16
2
1
0
0.005 0.007 0.009 0.011 0.013 0.015
MRR(cm3/s)
30
1 CO2 Fossil due to the machining process and machine base
CO2 Fossil (kg)
load
20 2 CO2 Fossil due to the dressing process
3 CO2 Fossil due to the coolant pump
4 CO2 Fossil due to the refill of the lost coolant (assumed
C
Fig.5.7Eco-efficiency evaluation of a selected grinding process (Studer S120) (Li etal. 2012)
5.3 Eco-efficiency of Grinding Processes 111
Overall when comparing the plotted surface, the CBN grinding wheel
shows much better capability for meeting high surface finishing requirements
(RZ<1.4m). This can be explained by the higher micro hardness of the abra-
sive grains of the CBN grinding wheel. In comparison, the use of Al2O3 grinding
wheel leads to significantly higher trend of CO2 Fossil due to the increased MRR.
This is caused by the higher wear and therefore the need for repeated dressing of
the grinding wheel.
5.4Summary
References
Borkowski JA, Szymaski A (1992) Uses of abrasives and abrasive tools. Ellis Horwood,
Chichester
Cecimo (2009) Concept description for CECIMOs self-regulatory initiative (SRI) for
the sector specific implementation of the directive 2005/32/EC. Available online.
http://www.ecodesign-info.eu/documents/machin_tools_va_20oct09.pdf last visited
08/02/2010
Clarens AF, Zimmerman JB, Keoleian GA, Hayes KF, Skerlos SJ (2008) Comparison of life
cycle emissions and energy consumption for environmentally adapted metalworking fluid
systems. Environ Sci Technol 42(22):85348540
Dettmer GT (2006) Nichtwassermischbare Khlschmierstoffe auf Basis nachwachsender
Rohstoffe. PhD thesis in Technischen Universitt Carolo-Wilhelmina zu Braunschweig,
Germany
Graham D (2000) Going dry. Manufact Eng 124(1):7278
112 5 Eco-efficiency of Manufacturing Processes
Greim J, Schwetz KA (2006) Boron carbide, boron nitride, and metal borides. Ullmanns
Encyclopedia of Industrial Chemistry, Wiley
Gu J, Barber G, Tung S, Gu R (1999) Tool life and wear mechanism of uncoated and coated mill-
ing inserts. Wear 225229(1):273284
International Energy Agency (IEA) (2011) CO2 emissions from fuel combustion high-
lights. Online publication. http://www.iea.org/co2highlights/co2highlights.pdf last visited
01/03/2012
Jawahir IS, Brinksmeier E, MSaoubi R, Aspinwall DK, Outeiro JC, Meyer D, Umbrello D,
Jayal AD (2011) Surface integrity in material removal processes: recent advances. CIRP Ann
Manufact Techonol 60(2):603626
Klocke F, Knig W (2005) Fertigungsverfahren 2Schleifen, Lppen. Springer Verlag GmbH,
Berlin Honen
Li W, Kara S (2015) Characterising energy efficiency of electrical discharge machining (EDM)
processes. Procedia CIRP 29:263268
Li W, Winter M, Kara S, Herrmann C (2012) Eco-efficiency of manufacturing processes: a grind-
ing case. CIRP AnnalsManufacturing Technology 61(1):5962
Marinescu ID, Tonshoff HK, Inasaki I (2000) Handbook of ceramic grinding and polishing.
William Andrew Publishing/Noyes, Norwich
Schiefer E (2000) kologische Bilanzierung von Bauteilen fr die Entwicklung umweltgerechter
Produkte am Beispiel spanender Fertigungsverfahren. PhD thesis in Technische Universitt
Darmstadt, Germany
Sreejith PS, Ngoi BKA (2000) Dry machining: machining of the future. J Mater Process Technol
101(1):287291
Stanford M, Lister PM (2002) The future role of metalworking fluids in metal cutting operations.
Ind Lubr Tribol 54(1):1119
Wentorf RH (1961) Synthesis of cubic form of boron nitride. J Chem Phys 34(3):809812
Winter M, hlschlger G, Dettmer T, Ibbotson S, Kara S, Herrmann C (2012) Using jatropha
oil based metalworking fluids in machining processes: a functional and ecological life cycle
evaluation. In: 19th CIRP international conference on life cycle engineering, Berkeley
Chapter 6
Implementation Towards Improving
Energy and Eco-efficiency
of Manufacturing Processes
The previous chapters have validated the results of the proposed methodology
for characterizing the energy and eco-efficiency of unit processes. The charac-
terization is around the development of unit process energy consumption models;
because, the model response-SEC is the indicator of energy efficiency; and the
associated environmental impacts of unit process are mainly due the consumption
of electricity. The results among all the tested processes proved the reliability of
the proposed methodology. However, due to the nature of empirical modelling, the
meaning of the model coefficients remains to be answered.
World Energy Council (WEC) suggests that Energy efficiency improvements
refer to a reduction in the energy used for a given service (heating, lighting, etc.)
or level of activity (WEC 2008). The understanding of energy efficiency should
evolve from a simple input/output ratio towards the global efforts for energy
reduction. Therefore, the derived SEC models need to be used for developing
energy efficiency strategies.
This chapter firstly explores the mechanism of model coefficients, by decom-
posing the SEC model. Then, a clustered model for material removal processes are
presented and discussed. The following section discusses the challenges of charac-
terizing energy efficiency in industries, and proposes a modified methodology with
a case study in a biomedical products and services company. The derived energy
consumption information is also compared with current LCI database in order to
improve its quality. Finally, the strategies for improving energy efficiency are pro-
posed from different perspectives.
Table6.1Comparison between c1 and fixed power for the tested machine tools
c1
Machine c1 (kW) Fixed power Pf (kW) Pf (%)
Colchester Tornado A50 2.191 1.16 189
Mori Seiki NL200MC/500 2.445 1.58 155
IKEGAI AX 20 4.415 1.77 249
Mori Seiki SL-15 2.273 1.48 154
Nakamura TMC-15 2.349 1.54 153
Mori Seiki Dura Vertical 5100 1.344 1.02 132
Fadal VMC 4020 1.330 0.74 180
DMG DMU 60P 5.863 5.45 108
Reinecker ISA 110 3.399 2.16 157
Studer S120 4.781 1.69 283
Studer S40 6.510 3.69 176
BOY15S 1.933 1.41 137
BOY 15 2.239 1.02 220
Battenfeld BA500CD 3.816 3.26 117
SODICK A3R 2.250 1.40 161
AGIE TRON 30L 2.111 1.26 168
6.1 Investigation into Model Coefficients 115
the process load is the amount of energy used for actual work. For the case of
turning process, the total energy consumption should be the sum of fixed energy
consumption and cutting energy at tool tip. However, this is an ideal situation.
During the process, frictions and vibrations would generate heat, which results in
a form of energy losses. In addition, the efficiency of the spindle drive motor and
transmission system under dynamic loads is another complex topic. In short, the
machine tools are complex systems; and the gap between empirical and theoretical
models is mainly due to the over simplification of the system. In order to further
understand the gap, model decomposition was conducted and presented in the fol-
lowing section.
6.1.2SEC Decomposition
A turning case was selected for SEC decomposition owing to the simplicity of
the process comparing to other tested manufacturing processes. The existing
knowledge about turning process would also benefit the investigation. Dietmair
and Verl (2008) used an approach in which they changed the operating condi-
tions of a milling machine tool state by state in order to develop the energy con-
sumption profile of each component. However, it is very difficult to differentiate
between the energy consumed by the spindle drive motor and that consumed by
the servo motors. It is also impractical to directly measure at the component level.
Alternatively, this research decomposed the energy consumption based on differ-
ent utilizations, and selected Colchester Tornado A50 to demonstrate the results.
When a machine is in stand-by mode or is idle, a fixed amount of power is
required by the auxiliary components and unloaded components to ensure that the
machine is ready for operation when required. In practice, an air cut is used to
check the numerical control program; in this approach the spindle rotates and the
cutting tool moves according to the program without contacting the workpiece.
Thus, due to these essential operations there is an additional power requirement
for an air cut compared to the fixed power required at stand-by mode. The power
gap between the normal cut and the air cut is the energy utilized to remove mate-
rial from the workpiece. However, this amount of energy is not the amount of
energy used by the tool tip for material removal. Due to friction effects, some of
this energy is converted into heat. Since the modelling response is SEC, the energy
consumption is decomposed based on the operating period required to remove
1cm3 of material. Therefore, the SEC can be separated into the following four
segments, as shown in Fig.6.1.
The specific fixed energy (SFE): this is the energy used to ensure that the machine
is ready over the operating period required to remove 1cm3 of the workpiece.
The specific operational energy (SOE): this is the energy used to enable the
essential operations such as spindle rotation and cutting tool movement over the
operating period required to remove 1cm3 of the workpiece.
116 6 Implementation Towards Improving Energy
8000
Material : Mild Steel 1020
0: +7
: 95
Unproductive
f: 0.2mm/rev Specific
Power
d: 1mm Unproductive
6000 v: 150m/min Energy
(SUE)
Effective Power (W)
Specific
Tool Tip
Power
Normal cut Tool tip
4000 Energy
(STE)
Operational
Specific
Air cut
Power
Operational
2000 Energy
(SOE)
Fixed Power
Stand- Stand- Specific Fixed
by by Energy
0
(SFE)
Time (s)
The specific tool tip energy (STE): this is the energy required by the tool tip to
remove 1cm3 of the workpiece.
The specific unproductive energy (SUE): this is the energy that is converted to
heat or other form rather than cutting energy over the operating period required
to remove 1cm3 of the workpiece.
The first two segments can be easily measured and separated from SEC based on
power measurements of air cuts.
For the Colchester Tornado A50, the measured fixed power was 1.16kW with a
standard deviation of 0.003kW. In the stand-by mode, the hydraulic pump motor
is used to maintain the pressure level of the clamping system and the servo motors
draw power to maintain their position. Also, the control system and other auxil-
iary components are activated to allow the machine to be ready for operation on
request. During the cutting operations, these components remain at the same status
except for the servo motors when extra cutting loads are allocated.
The machine tool requires supplementary power to perform the essential opera-
tions associated with the cutting process, such as rotating the spindle and mov-
ing the cutting tool along the programmed cutting path, which can be measured
through the air cuts. Since the servo motors that control the X and Y axes motions
are already enabled during the stand-by mode, the additional power required to
move the cutting tool is insignificant. Hence, the required operational power
6.1 Investigation into Model Coefficients 117
1.5
Poperational = 0.0003RPM - 0.0038
Effective Power (kW) R-square = 0.9819
0.5
0 RPM
0 1000 2000 3000 4000
is mainly due to the spindle drive motor, which is closely related to the rotation
speed of the spindle. Since the power consumption of air cuts was additionally
recorded, the SOE can be easily calculated and separated from SEC. As the tested
machine tool has a relative small spindle drive motor (3.5kW/4.7HP), the opera-
tional power for spindle rotation only accounts for a small proportion of the total
power demand, which follows a linear relationship with the rotational speed as
shown in Fig.6.2.
However, the other two segments, STE and SUE cannot be separated directly.
As discussed in Sect.2.1, the tool tip energy consumption for turning processes
can be estimated based on cutting forces. The comprehensive information avail-
able about tool geometries, material properties, and the formulas provided by
SECO Tools are used in this research (SECO Tools 2009). The cutting forces
can be obtained, according to Eq.6.1, in terms of the specific cutting force kc, the
feed, and depth of cut. The specific cutting force is a function of the shear stress
of the workpiece material (kc1.1), the geometric properties of the cutting action
(approach angle , rake angle 0), and material exponent (mc).
1 0.010
Fc = kc f d = kc1.1 f d (6.1)
(f sin )mc
In order to evaluate the accuracy of this equation the cutting forces were measured
under different cutting conditions using a Kistler Dynamometer. The observed
cutting forces are compared with predicted values in Fig.6.3. As Eq.6.1 suggests,
the shear zone, which is the product of the feed rate and depth of cut, has a lin-
ear relationship with the cutting force. Thus, the cutting forces are plotted against
the share zone. While the accuracy of the predicted cutting force for the removal
of aluminum is relatively high, the reliability decreases when harder materials are
118 6 Implementation Towards Improving Energy
600
400
200
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Shear Zone (mm)
Fig.6.3Plot of measured cutting forces versus predicted cutting forces (Colchester Tornado A50)
considered. Thus, the cutting force model can only provide a rough estimate of the
energy requirements of tool tip for material removal.
The unproductive energy is even more difficult to estimate. From a thermo-
dynamic perspective the machine normally generates more heat than the levels
expected due to the material removal process. The friction created at the bear-
ing unit and transmission system could be additional sources of the unproductive
energy (Altintas 2000). Jedrzejewski etal. (2005) suggested that the unproductive
power in a machine system is due to a wide range of factors which result in a
complex, dynamic, and non-linear state for the thermal equilibrium that is charac-
terized by continuous changes in power losses, temperatures, and displacements.
They developed a hybrid model which was combined with the finite element
method and finite difference method to analyze the thermal performance of spin-
dle and bearing units. Abele etal. (2010) presented a comprehensive review about
spindle units including existing thermal models and they highlighted that the com-
plexity and inherent limitations of these models make it unlikely that the reliance
on experimental measurements will change in the near future. Direct measure-
ment of generated heat is also inapplicable in this case. As an alternative approach
this research uses the observed SEC values and subtracts the other segments SFE,
SOE, and STE from SEC to estimate SUE. All the measured or estimated data has
been processed with curve-estimation to characterize the relationship between
MRR and each segment of SEC. Therefore, the SEC model decomposition was
achieved. Figure6.4 shows the results in a selected MRR range (01cm3/s).
6.1 Investigation into Model Coefficients 119
SEC (kJ/cm3)
Unproductive
Power Specific
Unproductiv
e Energy
(SUE)
Specific
Tool Tip
Power
Tool tip
Energy
(STE)
Operational
Specific
Power
Operational
Energy
(SOE)
Fixed Power
Specific
Fixed
Energy
(SFE)
Fig.6.4Results of SEC decomposition for the Colchester Tornado A50 (Li and Kara 2011)
6.1.3Coefficients c0 and c1
The above presented analysis shows the difficulty of estimating energy consump-
tion of a machine tool from mechanical theories. Even obtaining the fixed power
used in stand-by mode is dependent on measurements. The friction-generated heat
and the energy converted to heat make the prediction of energy consumption more
complicated and less applicable. More importantly, the total energy consumption
of a machine tool is not only due to energy required for machine readiness and by
the tool tip. The essential operations and friction effects also consume a proportion
of the total energy. Thus, the coefficients c0 and c1 are not as the specific cutting
energy and idle power suggested by the exergy framework (Gutowski etal. 2007).
Since the empirical approach considers the machine tool to be a single holistic
system, it is difficult to precisely assign the factors or reasons for each coefficient
in the model. However, the decomposition of SEC has shown several trends for
each segment of energy consumption, which leads toward a qualitative definition
of the factors in relation to coefficient c0 and c1.
120 6 Implementation Towards Improving Energy
As shown in Fig.6.4, the empirical models for each segment of the energy con-
sumption can be approximated to Eqs.6.26.5.
1.16
SFE = (6.2)
MRR
0.196
SOE = 0.425 + (6.3)
MRR
0.089
STE = 1.522 + (6.4)
MRR
0.746
SUE = 0.452 + (6.5)
MRR
The fixed power consumption is the major components for SEC according to
Fig.6.4. The equation for SFE (see Eq.6.2) also suggests that the fixed power is
the main contributor for coefficient c1. An investigation into fixed power consump-
tion and the induced improvements would significantly reduce energy consump-
tion during both stand-by and processing periods.
The SOE only counts a small proportion of the total SFE in the tested case.
Since this share of energy consumption is mainly due to rotation of spindle drive
motor, the size and the type of the motor would affect the proportion of SOE.
Generally, the bigger the motor size the higher share of SOE. However, there are
exceptions to this linear trend owing to the efficiency of the motors. Thus, the
spindle motor characteristics affects the SOE so as the coefficient c0.
The share of STE mainly contributes to the coefficient c0 in SEC models. The
tool tip energy mainly refers to the cutting energy during machining. Previously
reviewed cutting force models have suggested that the involved factors include
material, tool geometry and cutting parameters. In other words, coefficient c0
is dynamic rather than a static value. However, a constant value for c0 has been
found adequate for the process mainly due to the trade-off between the operational
and tool tip energy consumption. For example, removing aluminum requires a rel-
atively low specific cutting force, but demands a high rotation speed resulting in a
comparatively high operational energy. Correspondingly, a constant value for c0
can still provide a reliable energy consumption prediction.
As shown in Fig.6.4, the trend of SUE suggests that the lower the cutting
load, the higher the unproductive energy. This refers to a close link between c1
for SEC model and the motor and transmission efficiency of the machine tool.
This behavior agrees with the feature of the induction motors, that the motor
efficiency is very low when rotating at a slow speed or with small loads. The
negative coefficient value in Eq.6.5 is most likely due to the errors of the esti-
mation. But the trend indicates that the unproductive energy become negligi-
ble when MRR reaches a certain point. This leads to the definition of a critical
region in which decreasing MRR results in an exponential increase of SEC, and
thus the region should be avoided if possible. In other words, the increase of MRR
6.1 Investigation into Model Coefficients 121
outside the critical region would not result in a significant reduction of total energy
consumption.
In summary, the coefficients c0 and c1 for the SEC models can be written as a
qualitative function of above mentioned factors, see in Eqs.6.6 and 6.7.
Chapter 4 presents the derived SEC models for a variety of machine tools and
manufacturing process. Although the values of the coefficients for SEC models
vary from machine to machine, they all agree with an inverse shape which ena-
bles the opportunity for model clustering and comparison. Since multiple mate-
rial removal processes have been investigated, the results for turning, milling and
grinding are further discussed.
The MRR features different range for different machining processes. For exam-
ple, the turning process can remove metal with a MRR from 0.1 to 15cm3/s for
finishing or roughing respectively; while, the MRR for grinding process is only
limited in the range of 0.0050.05cm3/s in order to achieve a high surface quality
(Kalpakjian and Schmid 2005). If the MRR is assumed to have extreme values, the
SEC models can be converted into Eqs.6.8 and 6.9.
c1
lim SEC = c0 + = c0 + = (6.8)
MRR0 0
c1
lim SEC = c0 + = c0 + 0 = c0 (6.9)
MRR
As above equation suggests, when MRR is very slow, the c0s contribution to SEC
becomes negligible. For instance, the MRR during grinding process is usually around
0.01cm3/s, resulting SEC above 1000kJ/cm3. Although the c0 values for tested
grinding process vary from 47 to 135, it accounts for less than 10% of the total SEC.
On contrary, when MRR exceeds certain point, the SEC only fluctuates within a lim-
ited range which can be approximate as a constant value. Therefore, it is possible to
cluster all the tested material removal processes into one model.
All the observed data has been combined into one file and processed with curve-
estimation, resulting in an inverse model with R-square value of 0.915, see in Eq.6.10.
5.180
SEC = 5.530 + (6.10)
MRR
122 6 Implementation Towards Improving Energy
Studer S40
Linear Region Reinecker ISA 110
25
Constant Region
5 6.654
Max=3.207
Min=1.319
2.147
1
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
MRR (cm3/s)
The negative value obtained for coefficient c0 is mainly due to the variety of tested
machine tools and the different ranges of MRR for different processes. As men-
tioned in Sect.6.1.3, this value can help to define a critical region in which reduc-
ing MRR would result in an exponential increase of SEC. According to Eq.6.10,
when MRR is less than 1cm3/s, the SEC becomes positive, and increases signifi-
cantly if further reducing MRR under 1cm3/s. Thus, a critical region of material
removal rate for material removal processes is 01cm3/s, which should be avoided
if possible.
Figure6.5 plots all the observed data together. The curve-estimation has been
repeated again for the data observed within the critical region. Besides the criti-
cal region, the trend of SEC turns to be constant when MRR is over 3cm3/s. This
region (MRR>3cm3) can be defined as a constant region, and thus the SEC can
be approximated to the average value of the observed data. In addition, the area
between critical and constant region can be simplified with a linear regression
model. Therefore, a clustered model has been derived for material removal pro-
cesses as shown in Eq.6.11.
2.165 + 4.489
MRR |0 < MRR 1
SEC = 8.908 2.254MRR |1 < MRR 3 (6.11)
2.147 |MRR > 3
6.2 Model Clustering for Metal Machining Processes 123
The clustered model is an ideal tool for rough estimation, since minimum input
information is required for energy consumption estimation. It also provides a big
picture over different machine tools and processes, which is helpful for obtaining
energy efficiency measures at a unit process level. However, owing to the variety
of the machine tools, the accuracy of the clustered model is not comparable with
those machine specific models derived in Chap.4. As shown in the linear region in
Fig. 6.5, the variability is relatively high among different machine tools and pro-
cesses. None of the 11 provided models by SPSS can achieve R-square value over
0.3. Although the error remains high when MRR varies from 1 to 3cm3/s, the SEC
can be limited in a relatively small range from 2.147 to 6.654kJ/cm3. More impor-
tantly, the clustered model can be safely used in the critical region and constant
region, which is adequate enough for estimation during the early design stage.
6.3.1Challenges in Industries
The other important aspect for model development is the possibility to change
process parameters. In industry, some machine tools are purpose built and run
under specific process conditions. In these cases, it is preferable to apply the
screening approach to obtain the energy profile of the machine tools. Conversely,
industries commonly use one machine tool to produce different parts; correspond-
ing, to different loads and cycle time result in different energy consumptions.
These processes provide the opportunity to develop energy consumption mod-
els regarding dynamic loads but the process parameters are generally configured
according to the quality requirements which limit the range of varying process
parameters. Testing the extreme cases also places machine tools at risk of costly
breakdowns. Therefore, the levels of variance for the process parameters need to
be defined properly in order to meet the industrial requirements as well as to pro-
vide a sufficient range for model development.
Apart from the technical constraints, it is essential to obtain the management
support. Ideally, the model development activities can be coupled with energy effi-
ciency projects, since the models provide valuable information. As presented in
Chap.4, the derived SEC models have revealed the important relationship between
the energy intensity of the process (i.e. energy consumption per unit volume/mass
of processed materials) and the productivity of the process (e.g. throughput rate
or material removal rate). The models also offer an accurate estimation and pre-
diction of energy consumptions, which benefit the energy accounting activities in
industries. Moreover, the energy efficiency aspect can be introduced to the stage
of configuring process parameters. It is also possible to reduce the energy con-
sumption whilst remaining at the same level of quality performances. Besides the
benefits, industry support can be enhanced by minimizing the experimental costs
and the interruption of the production. Therefore, the proposed methodology in
Chap.3 needs to be modified to meet industrial requirements.
As suggested by Montgomery, DoE is the critical stage for any experimental work,
but it is also time- and resource-consuming to filter out the insignificant factors
(2009). The presented cases in Chap.4 have proved that the production rate (e.g.
material removal rate or MRR, throughput rate or ) is the decisive factor for the
SEC. In fact, both MRR and throughput rate are dependent variables. For example,
the MRR of a turning process is the product of cutting speed, feed rate and depth
of cut. The types of raw materials also affect the selection of MRR. Obviously,
these independent factors have significant impacts on the energy consumption, and
are normally classified as design factors during the DoE stage. However, it is a
tedious procedure to statistically prove the significance or insignificance of each
factor on energy consumption. Alternatively, the conclusion obtained from pre-
vious works can be adapted for the similar manufacturing processes. Instead of
6.3 Implementation the Methodology of Energy Efficiency Evaluation 125
testing the significance of each factor on energy consumption, the qualitative rela-
tionship between process parameter and production rate can be easily observed.
For instance, drilling is similar to other conventional machining processes, such
as turning and milling; and, the MRR is a function of driller diameter, feed rate
and RPM (rotation per minute); other factors like workpiece materials and driller
materials have an indirect impact on the MRR; so, these factors can be directly
concluded as design factors for model development. As a result, the number of
experiments can be reduced considerably.
As mentioned in Sect.6.3.1, other issues need to be taken into account for
experiment planning. The first question which needs to be answered is whether
this process is suitable for empirical modelling or not. If all the process parame-
ters of one process remain constant, a screening approach is more suitable for that
process. Otherwise, the process can be considered for empirical modelling. Then,
each factor is closely observed in conjunction with the production rate. If one fac-
tor has a significant impact on production rate, that factor needs to be targeted as
design factor. Afterwards, the operational constraints need to be considered, which
leads to an addition planning step prior to all the modelling work. Cost analysis,
initial setup, scheduling issues and management support needs to be obtained dur-
ing this step. Once the experiments are approved by industry, the remaining stages
are similar to the original empirical approach, including physical experiments, sta-
tistical analysis and model validation. Therefore, the methodology can be modified
as shown in Fig.6.6.
The presented case is derived from a biomedical products and services company in
Australia. This company is continuously looking ways to improve their processes
with environmental awareness. The company has strategic targets for substan-
tially reducing energy and resource consumption as well as reducing its carbon
foot-print. To achieve that goal, the company has invested on improving the trans-
parency of its energy flows. Projects were formed to develop energy consumption
models or to obtain energy profiles at unit process level.
The manufacturing plant in western Sydney specializes in producing sterilized
IV and renal dialysis fluids. One of the key processes is the sheet extrusion line,
which produces plastic sheeting to construct containers for IV solutions. A typical
sheet extrusion line consists of an extruder, die, three roller stack, conveyor, pull
rollers, and a winder. In general, the plastic pallets or powders are firstly melted
and forced through the die by a screw. As the film is drawn out of the die, it passes
through the roller stack which controls the cooling rate, final thickness and the sur-
face of the sheet. The pull rollers are set to provide a uniform pressure, ensuring
that the plastic does not slip between the rollers. At the end of the sheet line, a
winder is normally used where the sheeting is stored.
126 6 Implementation Towards Improving Energy
Fig.6.6Flow chart of
the modified methodology List all the process factors
for characterizing energy
efficiency in industries
(Li etal. 2013) No Screening
variable?
approach
Yes
Empirical modelling
approach
Group as held-
Affecting No constant
production factors or
rate? nuisance
factors
Yes
No Modify the
Approved by
experiment
Manager?
plan
Yes
Conduct experiments
Statistical Analysis
Model Validation
Design of Experiments
According to Fig.6.6, a complete list of factors was first identified for this plastic
sheet extrusion line. The factors were categorized into four groups as shown in
Fig. 6.7. Some factors, such as air humidity and temperature, can be concluded
as uncontrollable factors. Other factors were further investigated in terms of their
variance and impacts on production rate.
6.3 Implementation the Methodology of Energy Efficiency Evaluation 127
Power
Contamination
Throughput
Material Dryness Air Humidity SEC
Properties Air Temperature
Material Environment
According to the production record, the tested sheet extrusion line produces
different types of plastic sheets. Two types of raw materials are used, PVC and
HDPE. The thickness of the plastic sheets also varies according to the size of the
IV bag, which requires different settings for the process parameters. Therefore, it
is useful and valuable to develop empirical models for this process.
The production rate of the process is measured by throughput rate, which is the
mass of produced plastic sheet per unit time. There are two methods to measure
the throughput of the tested process. The first method is to use the sensor which
gauges the thickness. The material mass flow can be calculated in conjunction
with the width and speed of the plastic sheet as well as the material intensity. The
second method calculates an average throughput rate, dividing the total mass of
one roll of plastic sheets by the production time. The first method is used to moni-
tor the throughput rate during production. Figure6.8 shows the throughput rate
trend during the start-up period (after changing the filter). The results suggest that
it requires at least 120s for the mass flow to stabilize.
Figure 6.8 also suggested that the change of throughput rate is due to the
increase of screw RPM from 6 to 100. The bottom roller RPM is adjusted accord-
ingly to maintain the level of line tension. According to the operators, the screw
RPM is normally set between 80 and 104. However, the theoretical range is much
greater from 0 to 120. The production schedule also suggests that once the RPM is
lower than 50, the demand cannot be met even if the process is running continu-
ously. Therefore, the screw RPM was targeted and grouped as a design factor, the
range of which is limited between 50 and 104.
Although different raw materials are used in production, the experiments with
HDPE were rejected by the plant management, since the scrap of HDPE is not
recycled onsite. Consequently, the tested material was limited to PVC, which is
cheaper than HDPE. The production of PVC sheets accounts for a considerably
128 6 Implementation Towards Improving Energy
higher proportion of the total output of this extrusion line. More importantly, the
PVC scraps are shredded and reused for extrusion, which resulted in minimal
material waste. Since the experiments were solely conducted with PVC, the tem-
peratures were also held constant. Therefore, the design factors were narrowed
down to the screw RPM.
In order to precisely characterize the relationship between power consumption
and screw RPM, it is important to test multiple levels of this single design fac-
tor, since 2-level experiments initial assume the relationship would be linear. After
discussion with the operators and plant manager, 6 levels of screw RPM would be
used in the experiment allowing an increment of 10 RPM between each step from
50 to 100.
As discussed in Sect.6.3.1, it is important to minimize the cost of experiments
as well as the interruption of the normal production. Figure6.8 suggests that the
machine takes at least 120s to stabilize. It was safe to allow 5min for process
stabilization, and another 10min for data collection. The mass throughput can be
manually recorded every 30s, so 10-min experiments offers 20 samples at each
level. The material cost was also estimated based on an average throughput rate.
Since the machine is running 6days a week, 24h per day, the experiments were
scheduled during one weekend, resulting in additional labor and operational cost.
In addition, electrician was also ordered to connect the metering device. The over-
all cost was under the project budget with the highest proportion due to labor
costs. Finally, the experiment plan was approved by the plant manager and the
project manager.
Experiment details
The tested extrusion sheet line is connected with different distribution boards
and these distribution boards supply other processes. Thus, the extrusion line was
separated into heating unit, screw unit, and the driving train unit (including the
rollers, conveyer and the winder) and three Chauvin Arnoux, C.A.8335 portable
power analyzers were used to meter these units simultaneously with 1s resolution.
Figure 6.9 shows the power consumption measurements of the screw unit. It
clearly indicates the steps of the experiments, including an unexpected filter
change during the process. The experiments steps were:
6.3 Implementation the Methodology of Energy Efficiency Evaluation 129
30000
RPM=80
25000
RPM=100
20000
RPM=90
Power (W)
15000
RPM=70
10000 Filter
Change
RPM=60
5000 Screw Power Consumption
RPM=50
Time (30 Seconds)
0
0 100 200 300 400 500 600
Once the process was at normal operating conditions the screw RPM was set to
50 and the drive train was set to keep the required tension on the line;
The throughput of the process was observed for 5min unit the process had
stabilized;
At the beginning of the experiment, the initial mass was measured. The through-
put was recorded every 30s for the duration of 10min;
After 10min had elapsed, the total mass was measured again to calculate the
average throughput rate;
The screw RPM was increased to 60 and the drive train was adjusted
accordingly;
Step 25 were repeated and the screw RPM was increased to 70, 80, 90 and 100.
Statistical analysis
A series of statistical analyses were conducted with a statistical package,
MiniTab version16.
The distributions of observed data at different levels were first analyzed. The
heating unit power curve showed a similar pattern over the entire experiment. Two
groups of measured heating power (screw RPM 50 and 100) were processed with
a two-sample t-test. The null hypothesis had a p-value of 1.000, which suggests
that both sets of observations belong to the same group and there is no variation
present in the population. Another t-test was conducted on the variances between
these two populations, which resulted in a p-value of 0.94. These two tests demon-
strated that the power consumption of the heating unit remains constant during the
operation. Therefore, the power consumption of the heating unit can be written as
Eq.6.12.
Pheating = 4.8699 + (6.12)
Unlike the power consumption of heating unit, other observed data shows an obvi-
ous difference when the screw RPM changes. At each level of screw RPM, the
130 6 Implementation Towards Improving Energy
data sets were found to follow a normal distribution. The means of all the observed
responses were listed in the Table6.2.
The observed responses were then processed using linear regression.
Figure6.10 shows the fitted linear model against screw RPM for power consump-
tion of screw unit, power consumption of the drive train unit, and throughput rate
() respectively. The models can be written as Eqs.6.136.15.
Pscrew = 1.479 + 0.2571 RPM + (6.13)
Fig.6.10Results of statistical analysis and model regression for the extrusion case. a Fitted lin-
ear model for screw power. b Fitted linear model for drive train power. c Fitted linear model for
throughput rate
132 6 Implementation Towards Improving Energy
2.35
SEC (kJ/g)
1.85
1.35
0.85
0.35
0 50 100
Throughput (g/s)
for unit turning and milling processes, covering different workpiece materials
(i.e. aluminum, cast iron, steel, etc.) and the purpose of machining (i.e. dressing,
roughing, or average). The LCA results encompass the direct electricity of the pro-
cess, the consumption of compressed air and lubricant oil, the removed metal, as
well as the machine and factory infrastructure, an example of unit turning process
is shown in Fig.6.12.
Since the derived SEC models target only electricity consumption, the model
results were compared with data in the LCI database. Table6.3 lists the compari-
son for turning processes with aluminum workpiece. The MRR was assumed for
each type of process, and both clustered model and a machine specific model for
Colchester Tornado A50 was compared with the LCI datasets.
As the table shows, the LCI datasets overestimate the electricity consumption
for turning 1kg of material. For some cases, the real process only uses one fifth
of the LCI suggested amount, which may result in a biased analysis and results.
Alternatively, the SEC models can be integrated with the LCI database. Both the
clustered model and machine specific models should be used, since the clustered
model can be extremely useful when only limited information is available. In that
way, the process parameters can be directly configured in the LCI and LCA analy-
sis. Thus, the quality of LCI database can be definitely improved.
134 6 Implementation Towards Improving Energy
From an operator point of view, a good energy efficiency measure should be easy-
to-practice, no hardware investment with short pay-back period. It mainly refers to
the activities relating to process and production planning and control. Based on the
derived SEC models and above model analysis, energy efficiency improvements
for manufacturing processes can be gained by either increasing production rate
(MRR or throughput) or reducing the fixed energy consumption.
For most of the manufacturing processes, operators have a certain degree of
freedom to select process parameters for producing a product. The constraints
mainly refer to the surface finishing, dimensional accuracy, process type, machine
tool capacity and tool life. Within the possible range, a higher material removal
rate or throughput is favored according to SEC models. For the case of turn-
ing, increasing MRR can be achieved by e.g. changing cutting parameters, alter-
ing cutting tools, applying cutting fluid, etc. In addition, critical region should
be avoided and prioritized for energy efficiency measures. For example, the
efforts for increasing MRR in process should be encouraged as the small amount
of MRR growth would result in a considerable energy reduction. However, changing
6.5 Energy Efficiency Strategies for Manufacturing Processes 135
14000
Vc= 200 m/min Vc= 200 m/min
f = 0.2 mm/revtooth f = 0.2 mm/revtooth
12000 d = 1.5 mm 12000 d = 1.5 mm
10000 10000
8000 8000
variable
6000 6000
4000 4000
variable fixed
2000 2000
process parameters also affects the surface finishing, dimensional accuracy, tool life,
etc. Therefore, the evaluation of eco-efficiency should be used. Continuing with the
case of grinding, the surface roughness, coolant consumption and the dressing pro-
cesses were integrated into the evaluation, as presented in Chap.5. The 3D figure
suggests a critical region for different grinding wheels (see in Fig.5.7). Within the
critical region, the benefits from increasing material removal rate become less sig-
nificant due to high requirement of surface roughness.
During the SEC decomposition, the fixed power consumption has been iden-
tified as a major contributor to coefficient c1. Figure6.13 compares an identical
milling process performed by two different milling machine tools. It clearly shows
that the difference of the total energy consumptions is mainly due to the fixed
parts, where DMG DMU 60P consumes 5 times of fixed energy consumption than
Mori Seiki Dura Vertical 5100 (5.45 and 1.02kW respectively). More importantly,
the fixed part not only accounts for a large proportion of the energy consumption
during processing period, but also accumulates total energy consumption over
other non-value adding periods. In industry, many machine tools are switched on
24/7, but only used for working in a relatively short period of time. Thus, inves-
tigation into the fixed energy consumption contains great potential for reducing
energy consumption of manufacturing processes.
From an operators point of view, the easiest approach for reducing fixed
energy consumption is to switch off the machine tool when the machine is idle.
However, this is not straight forward, as the fixed power ensures the machine read-
iness for operation and it takes time to regain the operational readiness. Therefore
the question of when to switch the machine off still remains unanswered.
136 6 Implementation Towards Improving Energy
8000
7000
Process Start
6000
Effective Power (W)
0
1 101 201 301 401 501 601 701
Time (s)
Table6.4List of start-up period and energy for the selected machine tools
Machine Start-up period (s) Start-up energy (kWh)
Colchester Tornado A50 30 0.004
Mori Seiki NL2000MC/500 150 0.038
Mori Seiki Dura Vertical 5100 110 0.021
DMG DMU 60P 100 0.096
Studer S120 110 0.036
Studer S40 250 0.256
BOY 15S (to 300C) 236 0.205
Battenfeld BA500CD (to 300C) 258 1.708
Figure 6.14 shows an entire power profile of the Mori Seiki Dura Vertical 5100
from power switch on to off.
As Fig.6.14 suggests, the machine requires over 150s in order to achieve
operational readiness. In general, if the machine tool is scheduled to be idle for
a period of time longer than start-up time, the machine needs to be switched off.
Table6.4 lists the start-up time and energy consumption for the selected machine
tools. There are three machine tools requiring long period for start-up. Studer S40
needs 240s to lubricate the driving system during machine start-up, while injec-
tion molding machines use around 4min to reach the desired heating temperature.
The high start-up energy demand for the Battenfeld BA500CD is due to the large
volume of melting material, and the high rated power of heating unit (8kW). In
that case, how temperature decreases during the idle stage should be investigated
as a future work in order to develop optimal switch-off strategies.
6.5 Energy Efficiency Strategies for Manufacturing Processes 137
The machine tool builders have become active to promote energy efficiency com-
ponents and machine tools. The energy consumption can be reduced by retrofitting
and replacing with energy efficient motors and pumps. Based on above analysis,
the fixed energy consumption is of particular interest in this research. Li etal.
(2011) attempted to assign the fixed power to each enabled components. But the
procedure has been found problematic due to the lack of information about each
component, as well as the accessibility to measure each motors. Alternatively, a
component-based description of fixed power consumption was derived by using
the power profile at different stage in conjunction with the wiring diagram of
the machine tool, as shown in Fig.6.15. The average fixed energy breakdown of
reviewed machine tools has suggested the hydraulic system and the cooling lubri-
cation system are the two hot-spots for energy efficiency improvements (see in
Fig.6.15d).
The new technologies such as variable speed hydraulic pumps are available in
market for improving the energy efficiency of manufacturing processes. The sav-
ings are believed to be considerable, but cost efficiency and pay-back periods need
to be taken into account as well.
Besides introducing new technologies, machine tool builders have a more
important role for supplying energy efficiency information about their prod-
ucts (Zein etal. 2011). In other words, the energy consumption of the machine
tools should be documented properly for the users. The proposed methodology
for deriving unit process energy consumption models can be implemented at the
machine tool builders side. Therefore, a list of energy consumption indicators
have been proposed to be included in the machine tool specifications (Table6.5).
The above list can be extended to include energy consumption of each compo-
nent, as well as their energy consumption trend against load and process param-
eters. This information is very helpful for selecting a machine tool from energy
efficiency perspective, as well as configuring a process chain in an energy efficient
way. For example, a shop floor has different milling and turning machine tools
featuring different capacities and energy consumption behaviors. With above pro-
posed indicators, the job can be allocated to a process which is capable but con-
sumes least amount of energy. The interrelationship among processes can be also
considered with a simulation tool (Herrmann etal. 2011). Owing to the dynamic
nature of SEC models, the quality of the simulation at a process chain level or sys-
tem level can be improved significantly.
Moreover, the provided information is also useful for the early stage of a prod-
uct life cycle. The product designers and process planners can design the product
and manufacturing processes in a way that reduces the energy consumption during
the manufacturing stage.
138 6 Implementation Towards Improving Energy
(a) (b)
9
Power (kW)
Effective
DMU 60P
6
3
Pfixed = 5.45
0
3
Mori Seiki
Power (kW)
Pfixed = 1.02
Effective
2 DuraVertical 5100
1
0
6
Mori Seiki Pfixed = 1.58
Power (kW)
Effective
4 NL2000MC/500
2
(c)
DMG DMU 60P 5.45
Mori Seiki Dura Auxiliary System
Vertical 5100 1.02
Mori Seiki Cooling & Lubrication
NL2000MC/500 1.58 System
Hydraulic System
Colchester Tornado
A50 1.16
Contol System
Studer S120 1.69 Servo Drives
(d)
Servo Auxiliary
Drives System
17% 19%
Control
System 6%
Cooling
Hydraulic Lubrication
System System
27% 31%
Fig.6.15Results of investigation into fixed energy consumption for the selected machine tools.
a Power profile. b Wiring scheme. c Component-based description of fixed power consumption.
d Average fixed energy breakdown of reviewed machine tools
6.5 Energy Efficiency Strategies for Manufacturing Processes 139
The design stage of a product life cycle is known to have a major influence on the
entire product life cycle, but it is difficult to control the impacts due to the absence
of physical objects (Stark 2011). Reliable models, simulations and other predicting
tools can certainly ease the difficulties. Based on the derived SEC models, there
are a few basic strategies for designers to reduce the energy consumption during
the manufacturing stage.
Material selection: the material properties have a direct impact on the produc-
tion rate during manufacturing processes. From the SEC point of view, mate-
rial with better machinability can significantly reduce the production time and
energy consumption.
Product shape: the size and geometric features of the products also affect the
manufacturing processes in terms of energy consumption. The large volume
of material removal are particularly inefficient from both energy and resource
point of view. If a forming or casting process is in front of a material removal
process, it is recommended to have formed or casted workpiece close to the end
shape, resulting in a minimum volume of material removal.
Surface finish and dimensional accuracy: the quality features of the product
are essential for fulfilling product functional requirements. Without losing the
surface finish or dimensional accuracy, a high MRR or throughput is preferred
which can be achieved by using different cutting tools, applying higher process
parameters, or efficient cutting fluid.
Although above three strategies may sound like common sense, the SEC models
can quantify the changes made during design stage, and enable proactive actions
to reducing the energy consumption during manufacturing stage. This can be
achieved by integrating SEC models with the design software such as Computer
Aided Design (CAD) software. Figure6.16 illustrates this concept, which consists
of a machine tool library, a CAD design interface, and an Energy consumption
module. The machine tool library is the database for the derived SEC models
for different machines and manufacturing processes. The CAD design interface
140 6 Implementation Towards Improving Energy
Quality requirements
Geometric
(surface roughness,
features
dimensional accuracy)
Total energy
Consumption
Module 3: Energy Consumption E
is the original environment in which the cutting volume of the workpiece can be
easily calculated. The energy consumption module uses the information from the
other two modules in conjunction with the selection of machine tools and process
parameters, and then calculates the energy consumption for producing this part
accordingly. This information can be fed back to the designers, which allows the
designers to compare different options regarding the energy consumption of manu-
facturing processes.
This concept can be enhanced by introducing other studies such as research
of surface integrity, tool wear, cost models, etc. For example, the design inter-
face requires a certain surface roughness; the models describing the relationship
process parameters and surface roughness would help the selection of process
parameters and MRR. The tool wear and other cost models can be further added to
provide economic information about the design. Therefore, future works are rec-
ommended to integrate unit process energy consumption models with other pro-
cess studies such as surface integrity and tool wear.
6.5 Energy Efficiency Strategies for Manufacturing Processes 141
6.6Summary
This chapter has further discussed the use of unit process energy consumption
models, besides the use for evaluating the energy and eco-efficiency of unit pro-
cesses. The derived SEC model has been both decomposed and clustered. The
SEC model decomposition has qualitatively suggested the factors related to the
model coefficients. The findings have resulted in the definition of critical regions,
in which unproductive energy consumption is high, and the SEC increases expo-
nentially along decreasing MRR. The clustered model for metal machining pro-
cesses has provided a useful tool for rough estimation disregarding the type of
the machine tools or the processes. It also quantified the critical region for those
processes, which is the MRR between 0 and 1cm3/s. Moreover, this chapter has
modified the methodology for characterizing energy efficiency of manufactur-
ing process in industries. The planning stage (DoE) plays a critical role for model
development, where industrial requirements and practicality need to be considered
with caution. An extrusion line was used to demonstrate the methodology. The
resultant model shows a great ability to accurately predict energy consumption in
this tested case.
The results of energy consumption models have been compared with the exist-
ing LCI database, which suggests a dynamic calculation rather than a constant
value to improve the quality of the database. The energy efficiency strategies have
been developed accordingly from different perspectives. The fixed energy con-
sumption has been of particular focus, since it was believed to have great poten-
tial for reducing energy consumption during both process and non-value adding
periods. Moreover, a list of energy consumption indicators has been proposed for
machine tool builders. The SEC model is suggested to be included in the machine
tool specifications. In addition, the energy consumption models can be integrated
with design software, which provide another aspect for decision making.
References
Abele E, Altintas Y, Brecher C (2010) Machine tool spindle units. CIRP Ann Manuf Technol
59(2):781802
Altintas Y (2000) Manufacturing automation: metal cutting mechanics, machine tool vibrations,
and CNC design. Cambridge University Press, New York
142 6 Implementation Towards Improving Energy
Dietmair A, Verl A (2008) Energy consumption modeling and optimization for production
machines. In: 2008 IEEE international conference on sustainable energy technologies,
ICSET, Singapore
Gutowski T, Dahmus J, Thiriez A, Branham M, Jones A (2007) A thermodynamic characteriza-
tion of manufacturing processes. In: IEEE international symposium on electronics and the
environment, Orlando, FL
Herrmann C, Thiede S, Kara S (2011) Energy oriented simulation of manufacturing systems
concept and application. CIRP Ann Manuf Technol 60(1):4548
Jedrzejewski J, Kowal Z, Kwasny W, Modrzycki W (2005) High speed precise machine tools
spindle units improving. J Mater Process Technol 162163:615621
Kalpakjian S, Schmid SR (2005) Manufacturing engineering and technology, 5th edn. Prentice
Hall, Upper Saddle River
Li W, Kara S (2011) An empirical model for predicting energy consumption of manufactur-
ing processes: a case of turning process. In: Proceedings of the Institution of Mechanical
Engineers, Part B. J Eng Manuf 225(9):16361646
Li W, Kara S, Kornfeld B (2013) Developing unit process models for predicting energy
consumption in industry: a case of extrusion line. Re-engineering Manufacturing for
Sustainability, pp 147152
Li W, Zein A, Kara S, Herrmann C (2011) An investigation into fixed energy consumption of
machine tools. In: 18th CIRP international conference on life cycle engineering, Braunschweig,
Germany
SECO Tools (2009) MN2009-turning. The machining navigator catalogue
Stark J (2011) Decision engineering: product lifecycle management: 21st century paradigm for
product realisation, 2nd edn. Springer, New York
World Energy Council (2008) Energy efficiency polices around the world: review and evalua-
tion. Online publication: http://www.worldenergy.org/publications/energy_efficiency_poli-
cies_around_the_world_review_and_evaluation/default.asp. Last visited 1 Mar 2012
Zein A, Li W, Herrmann C, Kara S (2011) Energy efficiency measures for the design and opera-
tion of machine tools: an axiomatic approach. In: 18th CIRP international conference on life
cycle engineering, Braunschweig, Germany
Chapter 7
Conclusions
7.1Concluding Remarks
As stated in the beginning of the book, the overall research objective is to develop
a reliable methodology for characterizing the efficiency of manufacturing pro-
cesses from both energy and ecological perspectives. The specific energy con-
sumption, SEC, has been identified as the key indicator for the energy efficiency
of unit processes. It refers to the total energy consumption of a machine tool for
producing 1 unit of material. Therefore, developing SEC models for unit processes
is essential for evaluating energy and eco-efficiency of manufacturing processes.
After a large number of experiments, an empirical approach for character-
izing the relationship between process parameters and energy consumption has
been successfully developed. This methodology has been implemented on differ-
ent machine tools and manufacturing processes, including turning, milling, grind-
ing, injection moulding processes and electrical discharge machining. For the case
of material removal processes, SEC is the total energy consumption of removing
1cm3 material, whereas in the case of material addition process (such as injection
moulding), SEC uses the weight of processed material instead of volumetric meas-
ures. Notably, the volumetric measures can be converted into the weight measures
through the use of the workpiece density. Thus, this model response is generic and
transferable among the tested manufacturing processes.
For each type of manufacturing process, different machine tools have been
selected for model development and evaluations. Thus, the derived SEC models
are machine specific. Without exception, all the derived models agree with one
universal shape, where the production rate (material removal rate, through put)
plays a decisive role in the model. Based on the statistical analysis, the model can
exclude the sole impact of other parameters, i.e. workpiece material, tool geom-
etry, etc. However, those factors are interrelated to the production rate. Hence,
minimal input information is needed for predicting the energy consumption. More
importantly, the statistical results and additional validation runs have proved the
high accuracy of the derived models. The difference between predicted and meas-
ured energy consumption remains under 10%.
In addition to the evaluation energy efficiency, the value and the associated
environmental impacts of the process have been discussed. Instead of using a mar-
ket value, the value of unit process is suggested to be measured by physical char-
acteristics, such as removal volume, processed weight, surface roughness, etc. The
selection of indicators should agree with the primary objective of the unit process.
A general classification has been made for the tested manufacturing processes,
such as material removal process, surface finishing process, and thermoplastic pro-
cesses. Grinding has been found as a special case, due to the high importance of
surface integrity. Thus, the surface roughness has been added for the evaluation of
eco-efficiency.
The analyses of environmental impacts of unit processes face limitations due
to the absence of information regarding tools, coolant, etc. Regardless, electricity
consumption is generally believed to be the main source of environmental impacts.
The inverse of SEC can be directly used for evaluating the eco-efficiency for the
case of turning, milling under dry machining environmental, as well as injection
moulding. Since surface roughness is compulsory for grinding case, a 3D curve
has been generated to describe the interrelationship among process parameters,
surface roughness and CO2 emissions.
The evaluations results have been further investigated to develop strategies for
improving the energy and eco-efficiency of manufacturing processes.
The model decomposition has revealed the complexity of machine tool system,
which cannot be explained by a theoretical model. It also qualitatively suggests
the factors related to the derived empirical models. Although the coefficients of
the models vary from machine to machine, a clustered model has been obtained
for a rough estimation, and for developing energy efficiency measures. In order to
implement the energy efficiency characterization in industries, a modified method-
ology has been developed. Due to the limited process flexibility and the costs of
running experiments, a minimal number of experiments shall be planned to cover
the practical range of process variances. The demonstrated case study of a sheet
extrusion line has proved the reliability of the derived energy consumption mod-
els. In addition, the derived energy consumption models can certainly improve the
quality of life cycle inventory (LCI) database. Both machine specific models and
clustered model can be used for the future environmental analysis of a product or
process.
The energy efficiency strategies have been proposed based on the derived
models and the following analysis from different perspectives. Both of the model
decomposition and clustering has suggested a critical region of material removal
rate between 0 and 1cm3/s, which should be avoided if possible. The fixed power,
which is the energy due to the use of auxiliary components to ensure machine
readiness, has been identified as the priority for improvements. A feasible switch-
off strategy and retrofitting priorities have been proposed, and the benefits can be
quantified based on the derived models. The machine tool builders should also
7.1 Concluding Remarks 145
7.2Research Contributions
Without a doubt, this is an on-going research topic. According to the findings from
this research, there are three main areas for future efforts.
The proposed methodology for evaluating energy and eco-efficiency of man-
ufacturing processes can be extended and implemented in other processes, such
as casting process, metal forming process. The proposed methodology should be
adapted and modified to suit those processes. For example, the value of a metal
forming process is certainly different than the machining or injection molding.
Thus, further efforts are required to evaluate the energy and eco-efficiency of other
manufacturing processes.
The SEC models have their own limitations, which only capture the energy
during the value adding period. The overall efficiency of a process chain or man-
ufacturing system can be improved by integrating unit process models with sys-
tem simulation tools. As a result, the unit process models deal with the dynamics
of each process, while upper level simulations can manage the interrelationship
among the unit processes.
146 7Conclusions
Sample Data
Model summary
R R square Adjusted R square Std. error of the
estimate
0.997 0.993 0.993 0.596
ANOVA
Sum of squares df Mean square F Sig.
Regression 9432.483 1 9432.483 26,552.083 0.000
Residual 62.878 177 0.355
Total 9495.362 178
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 2.191 0.013 0.997 162.948 0.000
(Constant) 1.495 0.065 23.070 0.000
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
Model summary
R R square Adjusted R square Std. error of the
estimate
0.963 0.927 0.926 1.541
ANOVA
Sum of squares df Mean square F Sig.
Regression 5327.040 1 5327.040 2244.321 0.000
Residual 422.494 178 2.374
Total 5749.534 179
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 2.445 0.052 0.963 47.374 0.000
(Constant) 3.600 0.175 20.596 0.000
Appendix 3 153
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
Model summary
R R square Adjusted R square Std. error of the
estimate
0.991 0.981 0.981 1.512
ANOVA
Sum of squares df Mean square F Sig.
Regression 15,646.335 1 15,646.335 6840.151 0.000
Residual 299.653 131 2.287
Total 15,945.988 132
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 4.415 0.053 0.991 82.705 0.000
(Constant) 2.093 0.193 10.832 0.000
154 Appendix 3
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 4324.108 1 4324.108 2227.515 0.000
Residual 275.654 142 1.941
Total 4599.762 143
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
3
1/MRR(cm /s) 2.273 0.048 0.970 47.197 0.000
(Constant) 2.378 0.168 14.130 0.000
Appendix 3 155
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 4433.744 1 4433.744 1806.164 0.000
Residual 336.305 137 2.455
Total 4770.049 138
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
3
1/MRR (cm /s) 2.349 0.055 0.964 42.499 0.000
(Constant) 3.730 0.197 18.975 0.000
156 Appendix 3
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 34,318.546 1 34,318.546 9844.944 0.000
Residual 1934.678 555 3.486
Total 36,253.223 555
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 1.344 0.014 0.973 99.222 0.000
(Constant) 2.830 0.106 26.816 0.000
Appendix 3 157
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 42,668.595 1 42,668.595 19,921.793 0.000
Residual 1261.523 589 2.142
Total 43,930.118 590
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 1.330 0.009 0.986 141.145 0.000
(Constant) 2.845 0.079 36.130 0.000
158 Appendix 3
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 659,710.361 1 659,710.361 184,086.087 0.000
Residual 2035.545 568 3.584
Total 661,745.905 569
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. Error Beta
1/MRR (cm3/s) 5.863 0.014 0.998 429.053 0.000
(Constant) 2.411 0.106 22.789 0.000
Appendix 3 159
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 60,196.323 1 60,196.323 18,116.539 0.000
Residual 1103.146 332 3.323
Total 61,299.469 333
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
3
1/MRR (cm /s) 2.018 0.015 0.991 134.598 0.000
(Constant) 2.955 0.134 22.122 0.000
160 Appendix 3
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 38,153.415 1 38,153.415 14,386.779 0.000
Residual 933.496 352 2.652
Total 39,086.911 353
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 1.397 0.012 0.988 119.945 0.000
(Constant) 3.078 0.114 27.086 0.000
Appendix 3 161
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 613,966.400 1 613,966.400 25,937.998 0.000
Residual 8000.642 338 21.671
Total 621,967.043 339
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 6.422 0.040 0.994 161.053 0.000
(Constant) 2.425 0.356 6.820 0.000
162 Appendix 3
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 6,739,931.668 1 6,739,931.668 3258.593 0.000
Residual 119,964.657 58 2068.356
Total 6,859,896.325 59
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 3.399 0.060 0.991 57.084 0.000
(Constant) 47.269 12.060 3.920 0.000
Appendix 3 163
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 15,526,649.95 1 15,526,649.95 17,335.020 0.000
Residual 58,219.272 65 895.681
Total 15,584,869.22 65
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 4.781 0.036 0.998 131.663 0.000
(Constant) 70.240 6.846 10.260 0.000
164 Appendix 3
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 24,619,331.12 1 24,619,331.12 4976.690 0.000
Residual 286,921.844 58 4946.928
Total 24,906,252.96 58
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 6.510 0.092 0.994 70.546 0.000
(Constant) 135.103 17.325 7.798 0.000
Appendix 3 165
Model Plots
SEC (kJ/cm3)
MRR (cm3/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 32,782.975 1 32,782.975 54,054.513 0.000
Residual 605.267 998 0.606
Total 33,388.242 998
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
3
1/MRR (cm /s) 1.933 0.008 0.991 232.496 0.000
(Constant) 3.552 0.049 72.268 0.000
166 Appendix 3
Model Plots
SEC (kJ/g)
Throughput (g/s)
ANOVA
Sum of squares df Mean square F Sig.
Regression 9.947 1 9.947 816.269 0.000
Residual 0.353 29 0.012
Total 10.301 30
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
1/MRR (cm3/s) 3.816 0.134 0.983 28.570 0.000
(Constant) 1.582 0.103 15.434 0.000
Appendix 3 167
Model Plots
SEC (kJ/g)
Throughput (g/s)
Machine: BOY 15
Model summary
R R square Adjusted R square Std. error of the estimate
0.980 0.961 0.960 0.903
ANOVA
Sum of squares df Mean square F Sig.
Regression 923.508 1 923.508 1133.699 0.000
Residual 37.471 46 0.815
Total 960.979 47
Coefficients
Unstandardized Standardized coefficients t Sig.
coefficients
B Std. error Beta
3
1/MRR (cm /s) 2.239 0.066 0.980 33.670 0.000
(Constant) 2.201 0.446 4.937 0.000
168 Appendix 3
Model Plots
SEC (kJ/g)
Throughput (g/s)
Appendix 4
Machine: BOY 15
Humidity, 27, 48, 69, 80, 126 Methodology, vii, 2, 3, 7, 9, 15, 19, 20, 23,
Hydraulic, 18, 46, 78, 116, 137, 139 24, 26, 36, 42, 57, 69, 123, 124, 137,
Hydrocarbons, 41 143, 145
Milling, 3, 6, 7, 11, 13, 39, 5861, 64, 69, 70,
99, 103, 108, 121, 135, 137, 143, 144
I Minimal Quantity Lubrication (MQL), 42
Improvement measures, 2, 16 Minimum energy, 6, 8, 24
Incinerator, 108 MiniTab, 30, 49, 83, 129
Indicator, vii, 3, 6, 10, 15, 19, 23, 38, 39, 41, Model, 24, 7, 8, 11, 12, 1519, 2327, 29,
42, 54, 82, 83, 113, 137, 143, 144 31, 3336, 43, 47, 48, 5357, 59,
Injection molding, 3, 4, 14, 18, 26, 33, 39, 45, 6266, 69, 7478, 8387, 9396, 100,
7880, 82, 84, 85, 95, 99, 101, 102, 101, 104, 109, 111, 113115, 118126,
111, 123, 136 130134, 139, 141, 143145, 151169
Inscribed Central Composite (CCI), 29 Model response, 25, 83
Insert, 8, 48, 49, 51, 81, 85, 100 Mold, 26, 39, 79, 81, 102, 107
Insert geometry, 46 Monitoring, 3032, 88, 123
Interrelationship, vii, 16, 35, 42, 100, 108, Morbidity, 40
111, 137, 144, 145 Mortality, 40
Ishikawa diagram, 49, 59, 60, 72, 82, 127 Motor, 11, 12, 30, 45, 59, 62, 116, 117
MRR, 54, 59, 88, 101, 104, 110, 118,
120123, 125, 134, 139
K Multiple-linear regression, 25, 35, 36, 54, 56,
Kinematic, 23, 73 62, 64, 77
L N
LabVIEW, 3133 Nozzle, 67, 70
Land use, 40
Lathe, 17, 26, 46, 47
LCAdatabase, 3 O
LCI database, 113 Oblique cutting, 7
Life Cycle Assessment (LCA), 19, 40, 41, Occupational exposure limit (OEL), 108
106, 108, 132, 133, 146 One-factor-at-a-time (OFAT), 2830
Life Cycle Costing (LCC), 19 Operational readiness, 30, 135, 136
Life Cycle Impact Assessment (LCIA), 41 Operational states, 13
Life Cycle Inventory (LCI), viii, 4, 19, 40, Orthogonal machining, 7
108, 132, 144 Oxleys model, 7, 24
Lubrication, 13, 42, 67, 71, 137 Ozone depleting substance, 16
M P
Machinability, 5, 46, 48, 49, 139 Parameter, 27, 34, 62, 70, 83, 90, 93, 123, 125
Machine, 2 PC, 31, 32
Machine tool, vii, 11, 12, 15, 18, 24, 26, Peak current, 89
27, 30, 31, 33, 40, 46, 54, 5659, 62, Periphery system, 16
6568, 73, 75, 78, 81, 84, 88, 91, 100, Phosphorus, 41
102, 108, 114117, 119124, 135137, Piston, 67
139141, 143, 144 Plastic, 78, 102, 125127
Market, 1, 6, 10, 38, 137, 144 Plasticization, 79, 83
Material removal rate (MRR), vii, 34, 51, Policy, 6
5457, 62, 63, 65, 70, 74, 83, 90, Porosity, 39
9295, 99, 109, 122 Power, 7, 1113, 17, 18, 27, 28, 31, 45, 47, 48,
Measurand, 30, 31, 33 51, 52, 54, 60, 63, 66, 73, 85, 86, 91,
Metering, 25, 3033, 91, 123, 145 115, 128, 130
178 Index
V W
Validation, 25, 57, 95, 125, 130, 132 Waste, 41, 123
Value, 8, 19, 25, 38, 99, 103 Water, 16, 41, 88, 102
Variables, 24, 25, 29, 49, 57, 59, 65 Wet cut, 66
Variance, 90, 93, 124, 129, 130 Workpiece, 6, 8, 27, 39, 45, 46, 48, 49, 51, 53,
Viscosity, 33, 80, 90
Visual Instruments, 33 5760, 66, 71, 88, 94, 99, 115, 125
Voltage, 31
Volume, 2, 33, 34, 42, 51, 54, 81, 82, 124